2011-07-07 08:21:25 +00:00
|
|
|
/*
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
* Copyright (c) 2012 The WebRTC project authors. All Rights Reserved.
|
2011-07-07 08:21:25 +00:00
|
|
|
*
|
|
|
|
|
* Use of this source code is governed by a BSD-style license
|
|
|
|
|
* that can be found in the LICENSE file in the root of the source
|
|
|
|
|
* tree. An additional intellectual property rights grant can be found
|
|
|
|
|
* in the file PATENTS. All contributing project authors may
|
|
|
|
|
* be found in the AUTHORS file in the root of the source tree.
|
|
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* encode.c
|
|
|
|
|
*
|
|
|
|
|
* This file contains definition of funtions for encoding.
|
|
|
|
|
* Decoding of upper-band, including 8-12 kHz, when the bandwidth is
|
|
|
|
|
* 0-12 kHz, and 8-16 kHz, when the bandwidth is 0-16 kHz.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
#include <stdlib.h>
|
|
|
|
|
#include <string.h>
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
#include <stdio.h>
|
2011-07-07 08:21:25 +00:00
|
|
|
|
2018-02-01 14:12:55 +01:00
|
|
|
#include "modules/audio_coding/codecs/isac/main/source/structs.h"
|
|
|
|
|
#include "modules/audio_coding/codecs/isac/main/source/codec.h"
|
|
|
|
|
#include "modules/audio_coding/codecs/isac/main/source/pitch_estimator.h"
|
|
|
|
|
#include "modules/audio_coding/codecs/isac/main/source/entropy_coding.h"
|
|
|
|
|
#include "modules/audio_coding/codecs/isac/main/source/arith_routines.h"
|
|
|
|
|
#include "modules/audio_coding/codecs/isac/main/source/pitch_gain_tables.h"
|
|
|
|
|
#include "modules/audio_coding/codecs/isac/main/source/pitch_lag_tables.h"
|
|
|
|
|
#include "modules/audio_coding/codecs/isac/main/source/spectrum_ar_model_tables.h"
|
|
|
|
|
#include "modules/audio_coding/codecs/isac/main/source/lpc_tables.h"
|
|
|
|
|
#include "modules/audio_coding/codecs/isac/main/source/lpc_analysis.h"
|
|
|
|
|
#include "modules/audio_coding/codecs/isac/main/source/bandwidth_estimator.h"
|
|
|
|
|
#include "modules/audio_coding/codecs/isac/main/source/lpc_shape_swb12_tables.h"
|
|
|
|
|
#include "modules/audio_coding/codecs/isac/main/source/lpc_shape_swb16_tables.h"
|
|
|
|
|
#include "modules/audio_coding/codecs/isac/main/source/lpc_gain_swb_tables.h"
|
2018-04-27 04:31:53 +02:00
|
|
|
#include "modules/audio_coding/codecs/isac/main/source/isac_vad.h"
|
|
|
|
|
#include "modules/audio_coding/codecs/isac/main/source/pitch_filter.h"
|
2011-07-07 08:21:25 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
#define UB_LOOKAHEAD 24
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
|
2011-07-07 08:21:25 +00:00
|
|
|
/*
|
|
|
|
|
Rate allocation tables of lower and upper-band bottleneck for
|
|
|
|
|
12kHz & 16kHz bandwidth.
|
|
|
|
|
|
|
|
|
|
12 kHz bandwidth
|
|
|
|
|
-----------------
|
|
|
|
|
The overall bottleneck of the coder is between 38 kbps and 45 kbps. We have
|
|
|
|
|
considered 7 enteries, uniformly distributed in this interval, i.e. 38,
|
|
|
|
|
39.17, 40.33, 41.5, 42.67, 43.83 and 45. For every entery, the lower-band
|
|
|
|
|
and the upper-band bottlenecks are specified in
|
2011-12-01 13:01:39 +00:00
|
|
|
'kLowerBandBitRate12' and 'kUpperBandBitRate12'
|
2011-07-07 08:21:25 +00:00
|
|
|
tables, respectively. E.g. the overall rate of 41.5 kbps corresponts to a
|
|
|
|
|
bottleneck of 31 kbps for lower-band and 27 kbps for upper-band. Given an
|
|
|
|
|
overall bottleneck of the codec, we use linear interpolation to get
|
|
|
|
|
lower-band and upper-band bottlenecks.
|
|
|
|
|
|
|
|
|
|
16 kHz bandwidth
|
|
|
|
|
-----------------
|
2011-12-01 13:01:39 +00:00
|
|
|
The overall bottleneck of the coder is between 50 kbps and 56 kbps. We have
|
2011-07-07 08:21:25 +00:00
|
|
|
considered 7 enteries, uniformly distributed in this interval, i.e. 50, 51.2,
|
|
|
|
|
52.4, 53.6, 54.8 and 56. For every entery, the lower-band and the upper-band
|
2011-12-01 13:01:39 +00:00
|
|
|
bottlenecks are specified in 'kLowerBandBitRate16' and
|
|
|
|
|
'kUpperBandBitRate16' tables, respectively. E.g. the overall rate
|
2011-07-07 08:21:25 +00:00
|
|
|
of 53.6 kbps corresponts to a bottleneck of 32 kbps for lower-band and 30
|
|
|
|
|
kbps for upper-band. Given an overall bottleneck of the codec, we use linear
|
|
|
|
|
interpolation to get lower-band and upper-band bottlenecks.
|
|
|
|
|
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
*/
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* 38 39.17 40.33 41.5 42.67 43.83 45 */
|
2013-04-09 00:28:06 +00:00
|
|
|
static const int16_t kLowerBandBitRate12[7] = {
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
29000, 30000, 30000, 31000, 31000, 32000, 32000 };
|
2013-04-09 00:28:06 +00:00
|
|
|
static const int16_t kUpperBandBitRate12[7] = {
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
25000, 25000, 27000, 27000, 29000, 29000, 32000 };
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* 50 51.2 52.4 53.6 54.8 56 */
|
2013-04-09 00:28:06 +00:00
|
|
|
static const int16_t kLowerBandBitRate16[6] = {
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
31000, 31000, 32000, 32000, 32000, 32000 };
|
2013-04-09 00:28:06 +00:00
|
|
|
static const int16_t kUpperBandBitRate16[6] = {
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
28000, 29000, 29000, 30000, 31000, 32000 };
|
2011-07-07 08:21:25 +00:00
|
|
|
|
|
|
|
|
/******************************************************************************
|
|
|
|
|
* WebRtcIsac_RateAllocation()
|
|
|
|
|
* Internal function to perform a rate-allocation for upper and lower-band,
|
|
|
|
|
* given a total rate.
|
|
|
|
|
*
|
|
|
|
|
* Input:
|
|
|
|
|
* - inRateBitPerSec : a total bottleneck in bits/sec.
|
|
|
|
|
*
|
|
|
|
|
* Output:
|
|
|
|
|
* - rateLBBitPerSec : a bottleneck allocated to the lower-band
|
|
|
|
|
* in bits/sec.
|
|
|
|
|
* - rateUBBitPerSec : a bottleneck allocated to the upper-band
|
|
|
|
|
* in bits/sec.
|
|
|
|
|
*
|
|
|
|
|
* Return value : 0 if rate allocation has been successful.
|
|
|
|
|
* -1 if failed to allocate rates.
|
|
|
|
|
*/
|
|
|
|
|
|
2013-04-09 00:28:06 +00:00
|
|
|
int16_t WebRtcIsac_RateAllocation(int32_t inRateBitPerSec,
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
double* rateLBBitPerSec,
|
|
|
|
|
double* rateUBBitPerSec,
|
|
|
|
|
enum ISACBandwidth* bandwidthKHz) {
|
2013-04-09 00:28:06 +00:00
|
|
|
int16_t idx;
|
2011-07-07 08:21:25 +00:00
|
|
|
double idxD;
|
|
|
|
|
double idxErr;
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
if (inRateBitPerSec < 38000) {
|
|
|
|
|
/* If the given overall bottleneck is less than 38000 then
|
|
|
|
|
* then codec has to operate in wideband mode, i.e. 8 kHz
|
|
|
|
|
* bandwidth. */
|
2013-04-09 00:28:06 +00:00
|
|
|
*rateLBBitPerSec = (int16_t)((inRateBitPerSec > 32000) ?
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
32000 : inRateBitPerSec);
|
2011-07-07 08:21:25 +00:00
|
|
|
*rateUBBitPerSec = 0;
|
|
|
|
|
*bandwidthKHz = isac8kHz;
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
} else if ((inRateBitPerSec >= 38000) && (inRateBitPerSec < 50000)) {
|
|
|
|
|
/* At a bottleneck between 38 and 50 kbps the codec is operating
|
|
|
|
|
* at 12 kHz bandwidth. Using xxxBandBitRate12[] to calculates
|
|
|
|
|
* upper/lower bottleneck */
|
|
|
|
|
|
|
|
|
|
/* Find the bottlenecks by linear interpolation,
|
|
|
|
|
* step is (45000 - 38000)/6.0 we use the inverse of it. */
|
2011-07-07 08:21:25 +00:00
|
|
|
const double stepSizeInv = 8.5714286e-4;
|
|
|
|
|
idxD = (inRateBitPerSec - 38000) * stepSizeInv;
|
2013-04-09 00:28:06 +00:00
|
|
|
idx = (idxD >= 6) ? 6 : ((int16_t)idxD);
|
2011-07-07 08:21:25 +00:00
|
|
|
idxErr = idxD - idx;
|
2011-12-01 13:01:39 +00:00
|
|
|
*rateLBBitPerSec = kLowerBandBitRate12[idx];
|
|
|
|
|
*rateUBBitPerSec = kUpperBandBitRate12[idx];
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
if (idx < 6) {
|
2013-04-09 00:28:06 +00:00
|
|
|
*rateLBBitPerSec += (int16_t)(
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
idxErr * (kLowerBandBitRate12[idx + 1] - kLowerBandBitRate12[idx]));
|
2013-04-09 00:28:06 +00:00
|
|
|
*rateUBBitPerSec += (int16_t)(
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
idxErr * (kUpperBandBitRate12[idx + 1] - kUpperBandBitRate12[idx]));
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
*bandwidthKHz = isac12kHz;
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
} else if ((inRateBitPerSec >= 50000) && (inRateBitPerSec <= 56000)) {
|
|
|
|
|
/* A bottleneck between 50 and 56 kbps corresponds to bandwidth
|
|
|
|
|
* of 16 kHz. Using xxxBandBitRate16[] to calculates
|
|
|
|
|
* upper/lower bottleneck. */
|
|
|
|
|
|
|
|
|
|
/* Find the bottlenecks by linear interpolation
|
|
|
|
|
* step is (56000 - 50000)/5 we use the inverse of it. */
|
2011-07-07 08:21:25 +00:00
|
|
|
const double stepSizeInv = 8.3333333e-4;
|
|
|
|
|
idxD = (inRateBitPerSec - 50000) * stepSizeInv;
|
2013-04-09 00:28:06 +00:00
|
|
|
idx = (idxD >= 5) ? 5 : ((int16_t)idxD);
|
2011-07-07 08:21:25 +00:00
|
|
|
idxErr = idxD - idx;
|
2011-12-01 13:01:39 +00:00
|
|
|
*rateLBBitPerSec = kLowerBandBitRate16[idx];
|
|
|
|
|
*rateUBBitPerSec = kUpperBandBitRate16[idx];
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
if (idx < 5) {
|
2013-04-09 00:28:06 +00:00
|
|
|
*rateLBBitPerSec += (int16_t)(idxErr *
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
(kLowerBandBitRate16[idx + 1] -
|
|
|
|
|
kLowerBandBitRate16[idx]));
|
2011-07-07 08:21:25 +00:00
|
|
|
|
2013-04-09 00:28:06 +00:00
|
|
|
*rateUBBitPerSec += (int16_t)(idxErr *
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
(kUpperBandBitRate16[idx + 1] -
|
|
|
|
|
kUpperBandBitRate16[idx]));
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
*bandwidthKHz = isac16kHz;
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
} else {
|
|
|
|
|
/* Out-of-range botlteneck value. */
|
2011-07-07 08:21:25 +00:00
|
|
|
return -1;
|
|
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* limit the values. */
|
|
|
|
|
*rateLBBitPerSec = (*rateLBBitPerSec > 32000) ? 32000 : *rateLBBitPerSec;
|
|
|
|
|
*rateUBBitPerSec = (*rateUBBitPerSec > 32000) ? 32000 : *rateUBBitPerSec;
|
2011-07-07 08:21:25 +00:00
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
void WebRtcIsac_ResetBitstream(Bitstr* bit_stream) {
|
|
|
|
|
bit_stream->W_upper = 0xFFFFFFFF;
|
|
|
|
|
bit_stream->stream_index = 0;
|
|
|
|
|
bit_stream->streamval = 0;
|
|
|
|
|
}
|
2011-07-07 08:21:25 +00:00
|
|
|
|
2015-06-16 00:02:34 +02:00
|
|
|
int WebRtcIsac_EncodeLb(const TransformTables* transform_tables,
|
|
|
|
|
float* in, ISACLBEncStruct* ISACencLB_obj,
|
2013-04-09 00:28:06 +00:00
|
|
|
int16_t codingMode,
|
|
|
|
|
int16_t bottleneckIndex) {
|
2011-07-07 08:21:25 +00:00
|
|
|
int stream_length = 0;
|
|
|
|
|
int err;
|
|
|
|
|
int k;
|
|
|
|
|
int iterCntr;
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
double lofilt_coef[(ORDERLO + 1)*SUBFRAMES];
|
|
|
|
|
double hifilt_coef[(ORDERHI + 1)*SUBFRAMES];
|
2011-07-07 08:21:25 +00:00
|
|
|
float LP[FRAMESAMPLES_HALF];
|
|
|
|
|
float HP[FRAMESAMPLES_HALF];
|
|
|
|
|
|
|
|
|
|
double LP_lookahead[FRAMESAMPLES_HALF];
|
|
|
|
|
double HP_lookahead[FRAMESAMPLES_HALF];
|
|
|
|
|
double LP_lookahead_pf[FRAMESAMPLES_HALF + QLOOKAHEAD];
|
|
|
|
|
double LPw[FRAMESAMPLES_HALF];
|
|
|
|
|
|
|
|
|
|
double HPw[FRAMESAMPLES_HALF];
|
|
|
|
|
double LPw_pf[FRAMESAMPLES_HALF];
|
2013-04-09 00:28:06 +00:00
|
|
|
int16_t fre[FRAMESAMPLES_HALF]; /* Q7 */
|
|
|
|
|
int16_t fim[FRAMESAMPLES_HALF]; /* Q7 */
|
2011-07-07 08:21:25 +00:00
|
|
|
|
|
|
|
|
double PitchLags[4];
|
|
|
|
|
double PitchGains[4];
|
2013-04-09 00:28:06 +00:00
|
|
|
int16_t PitchGains_Q12[4];
|
|
|
|
|
int16_t AvgPitchGain_Q12;
|
2011-07-07 08:21:25 +00:00
|
|
|
|
|
|
|
|
int frame_mode; /* 0 for 30ms, 1 for 60ms */
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
int status = 0;
|
|
|
|
|
int my_index;
|
2011-07-07 08:21:25 +00:00
|
|
|
transcode_obj transcodingParam;
|
|
|
|
|
double bytesLeftSpecCoding;
|
2013-04-09 00:28:06 +00:00
|
|
|
uint16_t payloadLimitBytes;
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Copy new frame-length and bottleneck rate only for the first 10 ms data */
|
2011-07-07 08:21:25 +00:00
|
|
|
if (ISACencLB_obj->buffer_index == 0) {
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Set the framelength for the next packet. */
|
2011-07-07 08:21:25 +00:00
|
|
|
ISACencLB_obj->current_framesamples = ISACencLB_obj->new_framelength;
|
|
|
|
|
}
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* 'frame_mode' is 0 (30 ms) or 1 (60 ms). */
|
|
|
|
|
frame_mode = ISACencLB_obj->current_framesamples / MAX_FRAMESAMPLES;
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* buffer speech samples (by 10ms packet) until the frame-length */
|
|
|
|
|
/* is reached (30 or 60 ms). */
|
|
|
|
|
/*****************************************************************/
|
2011-07-07 08:21:25 +00:00
|
|
|
|
|
|
|
|
/* fill the buffer with 10ms input data */
|
|
|
|
|
for (k = 0; k < FRAMESAMPLES_10ms; k++) {
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
ISACencLB_obj->data_buffer_float[k + ISACencLB_obj->buffer_index] = in[k];
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* If buffersize is not equal to current framesize then increase index
|
|
|
|
|
* and return. We do no encoding untill we have enough audio. */
|
|
|
|
|
if (ISACencLB_obj->buffer_index + FRAMESAMPLES_10ms != FRAMESAMPLES) {
|
2011-07-07 08:21:25 +00:00
|
|
|
ISACencLB_obj->buffer_index += FRAMESAMPLES_10ms;
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* If buffer reached the right size, reset index and continue with
|
|
|
|
|
* encoding the frame. */
|
2011-07-07 08:21:25 +00:00
|
|
|
ISACencLB_obj->buffer_index = 0;
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* End of buffer function. */
|
2011-07-07 08:21:25 +00:00
|
|
|
/**************************/
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Encoding */
|
2011-07-07 08:21:25 +00:00
|
|
|
/************/
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
if (frame_mode == 0 || ISACencLB_obj->frame_nb == 0) {
|
|
|
|
|
/* This is to avoid Linux warnings until we change 'int' to 'Word32'
|
|
|
|
|
* at all places. */
|
2011-07-07 08:21:25 +00:00
|
|
|
int intVar;
|
|
|
|
|
/* reset bitstream */
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
WebRtcIsac_ResetBitstream(&(ISACencLB_obj->bitstr_obj));
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
if ((codingMode == 0) && (frame_mode == 0) &&
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
(ISACencLB_obj->enforceFrameSize == 0)) {
|
|
|
|
|
ISACencLB_obj->new_framelength = WebRtcIsac_GetNewFrameLength(
|
|
|
|
|
ISACencLB_obj->bottleneck, ISACencLB_obj->current_framesamples);
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
ISACencLB_obj->s2nr = WebRtcIsac_GetSnr(
|
|
|
|
|
ISACencLB_obj->bottleneck, ISACencLB_obj->current_framesamples);
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Encode frame length. */
|
2011-07-07 08:21:25 +00:00
|
|
|
status = WebRtcIsac_EncodeFrameLen(
|
|
|
|
|
ISACencLB_obj->current_framesamples, &ISACencLB_obj->bitstr_obj);
|
|
|
|
|
if (status < 0) {
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Wrong frame size. */
|
2011-07-07 08:21:25 +00:00
|
|
|
return status;
|
|
|
|
|
}
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Save framelength for multiple packets memory. */
|
2011-07-07 08:21:25 +00:00
|
|
|
ISACencLB_obj->SaveEnc_obj.framelength =
|
|
|
|
|
ISACencLB_obj->current_framesamples;
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* To be used for Redundant Coding. */
|
2011-07-07 08:21:25 +00:00
|
|
|
ISACencLB_obj->lastBWIdx = bottleneckIndex;
|
|
|
|
|
intVar = (int)bottleneckIndex;
|
|
|
|
|
WebRtcIsac_EncodeReceiveBw(&intVar, &ISACencLB_obj->bitstr_obj);
|
|
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Split signal in two bands. */
|
2011-07-07 08:21:25 +00:00
|
|
|
WebRtcIsac_SplitAndFilterFloat(ISACencLB_obj->data_buffer_float, LP, HP,
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
LP_lookahead, HP_lookahead,
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
&ISACencLB_obj->prefiltbankstr_obj);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
|
|
|
|
/* estimate pitch parameters and pitch-filter lookahead signal */
|
|
|
|
|
WebRtcIsac_PitchAnalysis(LP_lookahead, LP_lookahead_pf,
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
&ISACencLB_obj->pitchanalysisstr_obj, PitchLags,
|
|
|
|
|
PitchGains);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Encode in FIX Q12. */
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Convert PitchGain to Fixed point. */
|
|
|
|
|
for (k = 0; k < PITCH_SUBFRAMES; k++) {
|
2013-04-09 00:28:06 +00:00
|
|
|
PitchGains_Q12[k] = (int16_t)(PitchGains[k] * 4096.0);
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Set where to store data in multiple packets memory. */
|
|
|
|
|
if (frame_mode == 0 || ISACencLB_obj->frame_nb == 0) {
|
2011-07-07 08:21:25 +00:00
|
|
|
ISACencLB_obj->SaveEnc_obj.startIdx = 0;
|
|
|
|
|
} else {
|
|
|
|
|
ISACencLB_obj->SaveEnc_obj.startIdx = 1;
|
|
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Quantize & encode pitch parameters. */
|
2011-07-07 08:21:25 +00:00
|
|
|
WebRtcIsac_EncodePitchGain(PitchGains_Q12, &ISACencLB_obj->bitstr_obj,
|
|
|
|
|
&ISACencLB_obj->SaveEnc_obj);
|
|
|
|
|
WebRtcIsac_EncodePitchLag(PitchLags, PitchGains_Q12,
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
&ISACencLB_obj->bitstr_obj,
|
|
|
|
|
&ISACencLB_obj->SaveEnc_obj);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
|
|
|
|
AvgPitchGain_Q12 = (PitchGains_Q12[0] + PitchGains_Q12[1] +
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
PitchGains_Q12[2] + PitchGains_Q12[3]) >> 2;
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Find coefficients for perceptual pre-filters. */
|
2011-07-07 08:21:25 +00:00
|
|
|
WebRtcIsac_GetLpcCoefLb(LP_lookahead_pf, HP_lookahead,
|
|
|
|
|
&ISACencLB_obj->maskfiltstr_obj, ISACencLB_obj->s2nr,
|
|
|
|
|
PitchGains_Q12, lofilt_coef, hifilt_coef);
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Code LPC model and shape - gains not quantized yet. */
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
WebRtcIsac_EncodeLpcLb(lofilt_coef, hifilt_coef, &ISACencLB_obj->bitstr_obj,
|
|
|
|
|
&ISACencLB_obj->SaveEnc_obj);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Convert PitchGains back to FLOAT for pitchfilter_pre. */
|
2011-07-07 08:21:25 +00:00
|
|
|
for (k = 0; k < 4; k++) {
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
PitchGains[k] = ((float)PitchGains_Q12[k]) / 4096;
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Store the state of arithmetic coder before coding LPC gains. */
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
transcodingParam.W_upper = ISACencLB_obj->bitstr_obj.W_upper;
|
2011-07-07 08:21:25 +00:00
|
|
|
transcodingParam.stream_index = ISACencLB_obj->bitstr_obj.stream_index;
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
transcodingParam.streamval = ISACencLB_obj->bitstr_obj.streamval;
|
|
|
|
|
transcodingParam.stream[0] =
|
|
|
|
|
ISACencLB_obj->bitstr_obj.stream[ISACencLB_obj->bitstr_obj.stream_index -
|
|
|
|
|
2];
|
|
|
|
|
transcodingParam.stream[1] =
|
|
|
|
|
ISACencLB_obj->bitstr_obj.stream[ISACencLB_obj->bitstr_obj.stream_index -
|
|
|
|
|
1];
|
|
|
|
|
transcodingParam.stream[2] =
|
|
|
|
|
ISACencLB_obj->bitstr_obj.stream[ISACencLB_obj->bitstr_obj.stream_index];
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Store LPC Gains before encoding them. */
|
|
|
|
|
for (k = 0; k < SUBFRAMES; k++) {
|
|
|
|
|
transcodingParam.loFiltGain[k] = lofilt_coef[(LPC_LOBAND_ORDER + 1) * k];
|
|
|
|
|
transcodingParam.hiFiltGain[k] = hifilt_coef[(LPC_HIBAND_ORDER + 1) * k];
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Code gains */
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
WebRtcIsac_EncodeLpcGainLb(lofilt_coef, hifilt_coef,
|
|
|
|
|
&ISACencLB_obj->bitstr_obj,
|
|
|
|
|
&ISACencLB_obj->SaveEnc_obj);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
|
|
|
|
/* Get the correct value for the payload limit and calculate the
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
* number of bytes left for coding the spectrum. */
|
|
|
|
|
if ((frame_mode == 1) && (ISACencLB_obj->frame_nb == 0)) {
|
2011-07-07 08:21:25 +00:00
|
|
|
/* It is a 60ms and we are in the first 30ms then the limit at
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
* this point should be half of the assigned value. */
|
2011-07-07 08:21:25 +00:00
|
|
|
payloadLimitBytes = ISACencLB_obj->payloadLimitBytes60 >> 1;
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
} else if (frame_mode == 0) {
|
2011-07-07 08:21:25 +00:00
|
|
|
/* It is a 30ms frame */
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Subract 3 because termination process may add 3 bytes. */
|
2011-07-07 08:21:25 +00:00
|
|
|
payloadLimitBytes = ISACencLB_obj->payloadLimitBytes30 - 3;
|
|
|
|
|
} else {
|
|
|
|
|
/* This is the second half of a 60ms frame. */
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Subract 3 because termination process may add 3 bytes. */
|
2011-07-07 08:21:25 +00:00
|
|
|
payloadLimitBytes = ISACencLB_obj->payloadLimitBytes60 - 3;
|
|
|
|
|
}
|
|
|
|
|
bytesLeftSpecCoding = payloadLimitBytes - transcodingParam.stream_index;
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Perceptual pre-filtering (using normalized lattice filter). */
|
|
|
|
|
/* Low-band filtering. */
|
2011-07-07 08:21:25 +00:00
|
|
|
WebRtcIsac_NormLatticeFilterMa(ORDERLO,
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
ISACencLB_obj->maskfiltstr_obj.PreStateLoF,
|
|
|
|
|
ISACencLB_obj->maskfiltstr_obj.PreStateLoG,
|
|
|
|
|
LP, lofilt_coef, LPw);
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* High-band filtering. */
|
2011-07-07 08:21:25 +00:00
|
|
|
WebRtcIsac_NormLatticeFilterMa(ORDERHI,
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
ISACencLB_obj->maskfiltstr_obj.PreStateHiF,
|
|
|
|
|
ISACencLB_obj->maskfiltstr_obj.PreStateHiG,
|
|
|
|
|
HP, hifilt_coef, HPw);
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Pitch filter. */
|
2011-07-07 08:21:25 +00:00
|
|
|
WebRtcIsac_PitchfilterPre(LPw, LPw_pf, &ISACencLB_obj->pitchfiltstr_obj,
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
PitchLags, PitchGains);
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Transform */
|
2015-06-16 00:02:34 +02:00
|
|
|
WebRtcIsac_Time2Spec(transform_tables,
|
|
|
|
|
LPw_pf, HPw, fre, fim, &ISACencLB_obj->fftstr_obj);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Save data for multiple packets memory. */
|
|
|
|
|
my_index = ISACencLB_obj->SaveEnc_obj.startIdx * FRAMESAMPLES_HALF;
|
|
|
|
|
memcpy(&ISACencLB_obj->SaveEnc_obj.fre[my_index], fre, sizeof(fre));
|
|
|
|
|
memcpy(&ISACencLB_obj->SaveEnc_obj.fim[my_index], fim, sizeof(fim));
|
|
|
|
|
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
ISACencLB_obj->SaveEnc_obj.AvgPitchGain[ISACencLB_obj->SaveEnc_obj.startIdx] =
|
|
|
|
|
AvgPitchGain_Q12;
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Quantization and loss-less coding. */
|
|
|
|
|
err = WebRtcIsac_EncodeSpec(fre, fim, AvgPitchGain_Q12, kIsacLowerBand,
|
|
|
|
|
&ISACencLB_obj->bitstr_obj);
|
2011-07-07 08:21:25 +00:00
|
|
|
if ((err < 0) && (err != -ISAC_DISALLOWED_BITSTREAM_LENGTH)) {
|
|
|
|
|
/* There has been an error but it was not too large payload
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
(we can cure too large payload). */
|
2011-07-07 08:21:25 +00:00
|
|
|
if (frame_mode == 1 && ISACencLB_obj->frame_nb == 1) {
|
|
|
|
|
/* If this is the second 30ms of a 60ms frame reset
|
|
|
|
|
this such that in the next call encoder starts fresh. */
|
|
|
|
|
ISACencLB_obj->frame_nb = 0;
|
|
|
|
|
}
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
iterCntr = 0;
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
while ((ISACencLB_obj->bitstr_obj.stream_index > payloadLimitBytes) ||
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
(err == -ISAC_DISALLOWED_BITSTREAM_LENGTH)) {
|
2011-07-07 08:21:25 +00:00
|
|
|
double bytesSpecCoderUsed;
|
|
|
|
|
double transcodeScale;
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
if (iterCntr >= MAX_PAYLOAD_LIMIT_ITERATION) {
|
2011-07-07 08:21:25 +00:00
|
|
|
/* We were not able to limit the payload size */
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
if ((frame_mode == 1) && (ISACencLB_obj->frame_nb == 0)) {
|
2011-07-07 08:21:25 +00:00
|
|
|
/* This was the first 30ms of a 60ms frame. Although
|
|
|
|
|
the payload is larger than it should be but we let
|
|
|
|
|
the second 30ms be encoded. Maybe together we
|
|
|
|
|
won't exceed the limit. */
|
|
|
|
|
ISACencLB_obj->frame_nb = 1;
|
|
|
|
|
return 0;
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
} else if ((frame_mode == 1) && (ISACencLB_obj->frame_nb == 1)) {
|
2011-07-07 08:21:25 +00:00
|
|
|
ISACencLB_obj->frame_nb = 0;
|
|
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
if (err != -ISAC_DISALLOWED_BITSTREAM_LENGTH) {
|
2011-07-07 08:21:25 +00:00
|
|
|
return -ISAC_PAYLOAD_LARGER_THAN_LIMIT;
|
|
|
|
|
} else {
|
|
|
|
|
return status;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
if (err == -ISAC_DISALLOWED_BITSTREAM_LENGTH) {
|
2011-07-07 08:21:25 +00:00
|
|
|
bytesSpecCoderUsed = STREAM_SIZE_MAX;
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Being conservative */
|
2011-07-07 08:21:25 +00:00
|
|
|
transcodeScale = bytesLeftSpecCoding / bytesSpecCoderUsed * 0.5;
|
|
|
|
|
} else {
|
|
|
|
|
bytesSpecCoderUsed = ISACencLB_obj->bitstr_obj.stream_index -
|
|
|
|
|
transcodingParam.stream_index;
|
|
|
|
|
transcodeScale = bytesLeftSpecCoding / bytesSpecCoderUsed;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* To be safe, we reduce the scale depending on
|
|
|
|
|
the number of iterations. */
|
|
|
|
|
transcodeScale *= (1.0 - (0.9 * (double)iterCntr /
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
(double)MAX_PAYLOAD_LIMIT_ITERATION));
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Scale the LPC Gains. */
|
2011-07-07 08:21:25 +00:00
|
|
|
for (k = 0; k < SUBFRAMES; k++) {
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
lofilt_coef[(LPC_LOBAND_ORDER + 1) * k] =
|
2011-07-07 08:21:25 +00:00
|
|
|
transcodingParam.loFiltGain[k] * transcodeScale;
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
hifilt_coef[(LPC_HIBAND_ORDER + 1) * k] =
|
2011-07-07 08:21:25 +00:00
|
|
|
transcodingParam.hiFiltGain[k] * transcodeScale;
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
transcodingParam.loFiltGain[k] = lofilt_coef[(LPC_LOBAND_ORDER + 1) * k];
|
|
|
|
|
transcodingParam.hiFiltGain[k] = hifilt_coef[(LPC_HIBAND_ORDER + 1) * k];
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Scale DFT coefficients. */
|
2011-07-07 08:21:25 +00:00
|
|
|
for (k = 0; k < FRAMESAMPLES_HALF; k++) {
|
2013-04-09 00:28:06 +00:00
|
|
|
fre[k] = (int16_t)(fre[k] * transcodeScale);
|
|
|
|
|
fim[k] = (int16_t)(fim[k] * transcodeScale);
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Save data for multiple packets memory. */
|
|
|
|
|
my_index = ISACencLB_obj->SaveEnc_obj.startIdx * FRAMESAMPLES_HALF;
|
|
|
|
|
memcpy(&ISACencLB_obj->SaveEnc_obj.fre[my_index], fre, sizeof(fre));
|
|
|
|
|
memcpy(&ISACencLB_obj->SaveEnc_obj.fim[my_index], fim, sizeof(fim));
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Re-store the state of arithmetic coder before coding LPC gains. */
|
2011-07-07 08:21:25 +00:00
|
|
|
ISACencLB_obj->bitstr_obj.W_upper = transcodingParam.W_upper;
|
|
|
|
|
ISACencLB_obj->bitstr_obj.stream_index = transcodingParam.stream_index;
|
|
|
|
|
ISACencLB_obj->bitstr_obj.streamval = transcodingParam.streamval;
|
|
|
|
|
ISACencLB_obj->bitstr_obj.stream[transcodingParam.stream_index - 2] =
|
|
|
|
|
transcodingParam.stream[0];
|
|
|
|
|
ISACencLB_obj->bitstr_obj.stream[transcodingParam.stream_index - 1] =
|
|
|
|
|
transcodingParam.stream[1];
|
|
|
|
|
ISACencLB_obj->bitstr_obj.stream[transcodingParam.stream_index] =
|
|
|
|
|
transcodingParam.stream[2];
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Code gains. */
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
WebRtcIsac_EncodeLpcGainLb(lofilt_coef, hifilt_coef,
|
|
|
|
|
&ISACencLB_obj->bitstr_obj,
|
|
|
|
|
&ISACencLB_obj->SaveEnc_obj);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Update the number of bytes left for encoding the spectrum. */
|
|
|
|
|
bytesLeftSpecCoding = payloadLimitBytes - transcodingParam.stream_index;
|
|
|
|
|
|
|
|
|
|
/* Encode the spectrum. */
|
|
|
|
|
err = WebRtcIsac_EncodeSpec(fre, fim, AvgPitchGain_Q12, kIsacLowerBand,
|
|
|
|
|
&ISACencLB_obj->bitstr_obj);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
if ((err < 0) && (err != -ISAC_DISALLOWED_BITSTREAM_LENGTH)) {
|
2011-07-07 08:21:25 +00:00
|
|
|
/* There has been an error but it was not too large
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
payload (we can cure too large payload). */
|
2011-07-07 08:21:25 +00:00
|
|
|
if (frame_mode == 1 && ISACencLB_obj->frame_nb == 1) {
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* If this is the second 30 ms of a 60 ms frame reset
|
2011-07-07 08:21:25 +00:00
|
|
|
this such that in the next call encoder starts fresh. */
|
|
|
|
|
ISACencLB_obj->frame_nb = 0;
|
|
|
|
|
}
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
iterCntr++;
|
|
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* If 60 ms frame-size and just processed the first 30 ms, */
|
|
|
|
|
/* go back to main function to buffer the other 30 ms speech frame. */
|
|
|
|
|
if (frame_mode == 1) {
|
|
|
|
|
if (ISACencLB_obj->frame_nb == 0) {
|
2011-07-07 08:21:25 +00:00
|
|
|
ISACencLB_obj->frame_nb = 1;
|
|
|
|
|
return 0;
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
} else if (ISACencLB_obj->frame_nb == 1) {
|
2011-07-07 08:21:25 +00:00
|
|
|
ISACencLB_obj->frame_nb = 0;
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Also update the frame-length for next packet,
|
|
|
|
|
in Adaptive mode only. */
|
|
|
|
|
if (codingMode == 0 && (ISACencLB_obj->enforceFrameSize == 0)) {
|
2011-07-07 08:21:25 +00:00
|
|
|
ISACencLB_obj->new_framelength =
|
|
|
|
|
WebRtcIsac_GetNewFrameLength(ISACencLB_obj->bottleneck,
|
|
|
|
|
ISACencLB_obj->current_framesamples);
|
|
|
|
|
}
|
|
|
|
|
}
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
} else {
|
2011-07-07 08:21:25 +00:00
|
|
|
ISACencLB_obj->frame_nb = 0;
|
|
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Complete arithmetic coding. */
|
2011-07-07 08:21:25 +00:00
|
|
|
stream_length = WebRtcIsac_EncTerminate(&ISACencLB_obj->bitstr_obj);
|
|
|
|
|
return stream_length;
|
|
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
static int LimitPayloadUb(ISACUBEncStruct* ISACencUB_obj,
|
2013-04-09 00:28:06 +00:00
|
|
|
uint16_t payloadLimitBytes,
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
double bytesLeftSpecCoding,
|
|
|
|
|
transcode_obj* transcodingParam,
|
2013-04-09 00:28:06 +00:00
|
|
|
int16_t* fre, int16_t* fim,
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
double* lpcGains, enum ISACBand band, int status) {
|
|
|
|
|
|
|
|
|
|
int iterCntr = 0;
|
|
|
|
|
int k;
|
|
|
|
|
double bytesSpecCoderUsed;
|
|
|
|
|
double transcodeScale;
|
2013-04-09 00:28:06 +00:00
|
|
|
const int16_t kAveragePitchGain = 0.0;
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
|
|
|
|
|
do {
|
|
|
|
|
if (iterCntr >= MAX_PAYLOAD_LIMIT_ITERATION) {
|
|
|
|
|
/* We were not able to limit the payload size. */
|
|
|
|
|
return -ISAC_PAYLOAD_LARGER_THAN_LIMIT;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (status == -ISAC_DISALLOWED_BITSTREAM_LENGTH) {
|
|
|
|
|
bytesSpecCoderUsed = STREAM_SIZE_MAX;
|
|
|
|
|
/* Being conservative. */
|
|
|
|
|
transcodeScale = bytesLeftSpecCoding / bytesSpecCoderUsed * 0.5;
|
|
|
|
|
} else {
|
|
|
|
|
bytesSpecCoderUsed = ISACencUB_obj->bitstr_obj.stream_index -
|
|
|
|
|
transcodingParam->stream_index;
|
|
|
|
|
transcodeScale = bytesLeftSpecCoding / bytesSpecCoderUsed;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* To be safe, we reduce the scale depending on the
|
|
|
|
|
number of iterations. */
|
|
|
|
|
transcodeScale *= (1.0 - (0.9 * (double)iterCntr /
|
|
|
|
|
(double)MAX_PAYLOAD_LIMIT_ITERATION));
|
|
|
|
|
|
|
|
|
|
/* Scale the LPC Gains. */
|
|
|
|
|
if (band == kIsacUpperBand16) {
|
|
|
|
|
/* Two sets of coefficients if 16 kHz. */
|
|
|
|
|
for (k = 0; k < SUBFRAMES; k++) {
|
|
|
|
|
transcodingParam->loFiltGain[k] *= transcodeScale;
|
|
|
|
|
transcodingParam->hiFiltGain[k] *= transcodeScale;
|
|
|
|
|
}
|
|
|
|
|
} else {
|
|
|
|
|
/* One sets of coefficients if 12 kHz. */
|
|
|
|
|
for (k = 0; k < SUBFRAMES; k++) {
|
|
|
|
|
transcodingParam->loFiltGain[k] *= transcodeScale;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Scale DFT coefficients. */
|
|
|
|
|
for (k = 0; k < FRAMESAMPLES_HALF; k++) {
|
2013-04-09 00:28:06 +00:00
|
|
|
fre[k] = (int16_t)(fre[k] * transcodeScale + 0.5);
|
|
|
|
|
fim[k] = (int16_t)(fim[k] * transcodeScale + 0.5);
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
}
|
|
|
|
|
/* Store FFT coefficients for multiple encoding. */
|
|
|
|
|
memcpy(ISACencUB_obj->SaveEnc_obj.realFFT, fre,
|
|
|
|
|
sizeof(ISACencUB_obj->SaveEnc_obj.realFFT));
|
|
|
|
|
memcpy(ISACencUB_obj->SaveEnc_obj.imagFFT, fim,
|
|
|
|
|
sizeof(ISACencUB_obj->SaveEnc_obj.imagFFT));
|
|
|
|
|
|
|
|
|
|
/* Store the state of arithmetic coder before coding LPC gains */
|
|
|
|
|
ISACencUB_obj->bitstr_obj.W_upper = transcodingParam->W_upper;
|
|
|
|
|
ISACencUB_obj->bitstr_obj.stream_index = transcodingParam->stream_index;
|
|
|
|
|
ISACencUB_obj->bitstr_obj.streamval = transcodingParam->streamval;
|
|
|
|
|
ISACencUB_obj->bitstr_obj.stream[transcodingParam->stream_index - 2] =
|
|
|
|
|
transcodingParam->stream[0];
|
|
|
|
|
ISACencUB_obj->bitstr_obj.stream[transcodingParam->stream_index - 1] =
|
|
|
|
|
transcodingParam->stream[1];
|
|
|
|
|
ISACencUB_obj->bitstr_obj.stream[transcodingParam->stream_index] =
|
|
|
|
|
transcodingParam->stream[2];
|
|
|
|
|
|
|
|
|
|
/* Store the gains for multiple encoding. */
|
|
|
|
|
memcpy(ISACencUB_obj->SaveEnc_obj.lpcGain, lpcGains,
|
|
|
|
|
SUBFRAMES * sizeof(double));
|
|
|
|
|
/* Entropy Code lpc-gains, indices are stored for a later use.*/
|
|
|
|
|
WebRtcIsac_EncodeLpcGainUb(transcodingParam->loFiltGain,
|
|
|
|
|
&ISACencUB_obj->bitstr_obj,
|
|
|
|
|
ISACencUB_obj->SaveEnc_obj.lpcGainIndex);
|
|
|
|
|
|
|
|
|
|
/* If 16kHz should do one more set. */
|
|
|
|
|
if (band == kIsacUpperBand16) {
|
|
|
|
|
/* Store the gains for multiple encoding. */
|
|
|
|
|
memcpy(&ISACencUB_obj->SaveEnc_obj.lpcGain[SUBFRAMES],
|
|
|
|
|
&lpcGains[SUBFRAMES], SUBFRAMES * sizeof(double));
|
|
|
|
|
/* Entropy Code lpc-gains, indices are stored for a later use.*/
|
|
|
|
|
WebRtcIsac_EncodeLpcGainUb(
|
|
|
|
|
transcodingParam->hiFiltGain, &ISACencUB_obj->bitstr_obj,
|
|
|
|
|
&ISACencUB_obj->SaveEnc_obj.lpcGainIndex[SUBFRAMES]);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Update the number of bytes left for encoding the spectrum. */
|
|
|
|
|
bytesLeftSpecCoding = payloadLimitBytes -
|
|
|
|
|
ISACencUB_obj->bitstr_obj.stream_index;
|
|
|
|
|
|
|
|
|
|
/* Save the bit-stream object at this point for FEC. */
|
|
|
|
|
memcpy(&ISACencUB_obj->SaveEnc_obj.bitStreamObj,
|
|
|
|
|
&ISACencUB_obj->bitstr_obj, sizeof(Bitstr));
|
|
|
|
|
|
|
|
|
|
/* Encode the spectrum. */
|
|
|
|
|
status = WebRtcIsac_EncodeSpec(fre, fim, kAveragePitchGain,
|
|
|
|
|
band, &ISACencUB_obj->bitstr_obj);
|
|
|
|
|
if ((status < 0) && (status != -ISAC_DISALLOWED_BITSTREAM_LENGTH)) {
|
|
|
|
|
/* There has been an error but it was not too large payload
|
|
|
|
|
(we can cure too large payload). */
|
|
|
|
|
return status;
|
|
|
|
|
}
|
|
|
|
|
iterCntr++;
|
|
|
|
|
} while ((ISACencUB_obj->bitstr_obj.stream_index > payloadLimitBytes) ||
|
|
|
|
|
(status == -ISAC_DISALLOWED_BITSTREAM_LENGTH));
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
2015-06-16 00:02:34 +02:00
|
|
|
int WebRtcIsac_EncodeUb16(const TransformTables* transform_tables,
|
|
|
|
|
float* in, ISACUBEncStruct* ISACencUB_obj,
|
2013-04-09 00:28:06 +00:00
|
|
|
int32_t jitterInfo) {
|
2011-07-07 08:21:25 +00:00
|
|
|
int err;
|
|
|
|
|
int k;
|
|
|
|
|
|
|
|
|
|
double lpcVecs[UB_LPC_ORDER * UB16_LPC_VEC_PER_FRAME];
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
double percepFilterParams[(1 + UB_LPC_ORDER) * (SUBFRAMES << 1) +
|
2011-07-07 08:21:25 +00:00
|
|
|
(1 + UB_LPC_ORDER)];
|
|
|
|
|
|
|
|
|
|
double LP_lookahead[FRAMESAMPLES];
|
2013-04-09 00:28:06 +00:00
|
|
|
int16_t fre[FRAMESAMPLES_HALF]; /* Q7 */
|
|
|
|
|
int16_t fim[FRAMESAMPLES_HALF]; /* Q7 */
|
2011-07-07 08:21:25 +00:00
|
|
|
|
|
|
|
|
int status = 0;
|
|
|
|
|
|
|
|
|
|
double varscale[2];
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
double corr[SUBFRAMES << 1][UB_LPC_ORDER + 1];
|
|
|
|
|
double lpcGains[SUBFRAMES << 1];
|
2011-07-07 08:21:25 +00:00
|
|
|
transcode_obj transcodingParam;
|
2013-04-09 00:28:06 +00:00
|
|
|
uint16_t payloadLimitBytes;
|
2011-07-07 08:21:25 +00:00
|
|
|
double s2nr;
|
2013-04-09 00:28:06 +00:00
|
|
|
const int16_t kAveragePitchGain = 0.0;
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
int bytesLeftSpecCoding;
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Buffer speech samples (by 10ms packet) until the frame-length is */
|
|
|
|
|
/* reached (30 ms). */
|
2011-07-07 08:21:25 +00:00
|
|
|
/*********************************************************************/
|
|
|
|
|
|
|
|
|
|
/* fill the buffer with 10ms input data */
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
memcpy(&ISACencUB_obj->data_buffer_float[ISACencUB_obj->buffer_index], in,
|
|
|
|
|
FRAMESAMPLES_10ms * sizeof(float));
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* If buffer size is not equal to current frame-size, and end of file is
|
|
|
|
|
* not reached yet, we don't do encoding unless we have the whole frame. */
|
2011-07-07 08:21:25 +00:00
|
|
|
if (ISACencUB_obj->buffer_index + FRAMESAMPLES_10ms < FRAMESAMPLES) {
|
|
|
|
|
ISACencUB_obj->buffer_index += FRAMESAMPLES_10ms;
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* End of buffer function. */
|
2011-07-07 08:21:25 +00:00
|
|
|
/**************************/
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Encoding */
|
2011-07-07 08:21:25 +00:00
|
|
|
/************/
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Reset bit-stream */
|
|
|
|
|
WebRtcIsac_ResetBitstream(&(ISACencUB_obj->bitstr_obj));
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Encoding of bandwidth information. */
|
2011-07-07 08:21:25 +00:00
|
|
|
WebRtcIsac_EncodeJitterInfo(jitterInfo, &ISACencUB_obj->bitstr_obj);
|
|
|
|
|
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
status = WebRtcIsac_EncodeBandwidth(isac16kHz, &ISACencUB_obj->bitstr_obj);
|
2011-07-07 08:21:25 +00:00
|
|
|
if (status < 0) {
|
|
|
|
|
return status;
|
|
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
s2nr = WebRtcIsac_GetSnr(ISACencUB_obj->bottleneck, FRAMESAMPLES);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
|
|
|
|
memcpy(lpcVecs, ISACencUB_obj->lastLPCVec, UB_LPC_ORDER * sizeof(double));
|
|
|
|
|
|
|
|
|
|
for (k = 0; k < FRAMESAMPLES; k++) {
|
|
|
|
|
LP_lookahead[k] = ISACencUB_obj->data_buffer_float[UB_LOOKAHEAD + k];
|
|
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Find coefficients for perceptual pre-filters. */
|
2011-07-07 08:21:25 +00:00
|
|
|
WebRtcIsac_GetLpcCoefUb(LP_lookahead, &ISACencUB_obj->maskfiltstr_obj,
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
&lpcVecs[UB_LPC_ORDER], corr, varscale, isac16kHz);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
|
|
|
|
memcpy(ISACencUB_obj->lastLPCVec,
|
|
|
|
|
&lpcVecs[(UB16_LPC_VEC_PER_FRAME - 1) * (UB_LPC_ORDER)],
|
|
|
|
|
sizeof(double) * UB_LPC_ORDER);
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Code LPC model and shape - gains not quantized yet. */
|
2011-07-07 08:21:25 +00:00
|
|
|
WebRtcIsac_EncodeLpcUB(lpcVecs, &ISACencUB_obj->bitstr_obj,
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
percepFilterParams, isac16kHz,
|
|
|
|
|
&ISACencUB_obj->SaveEnc_obj);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* the first set of lpc parameters are from the last sub-frame of
|
|
|
|
|
* the previous frame. so we don't care about them. */
|
2011-07-07 08:21:25 +00:00
|
|
|
WebRtcIsac_GetLpcGain(s2nr, &percepFilterParams[UB_LPC_ORDER + 1],
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
(SUBFRAMES << 1), lpcGains, corr, varscale);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
|
|
|
|
/* Store the state of arithmetic coder before coding LPC gains */
|
|
|
|
|
transcodingParam.stream_index = ISACencUB_obj->bitstr_obj.stream_index;
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
transcodingParam.W_upper = ISACencUB_obj->bitstr_obj.W_upper;
|
|
|
|
|
transcodingParam.streamval = ISACencUB_obj->bitstr_obj.streamval;
|
|
|
|
|
transcodingParam.stream[0] =
|
|
|
|
|
ISACencUB_obj->bitstr_obj.stream[ISACencUB_obj->bitstr_obj.stream_index -
|
|
|
|
|
2];
|
|
|
|
|
transcodingParam.stream[1] =
|
|
|
|
|
ISACencUB_obj->bitstr_obj.stream[ISACencUB_obj->bitstr_obj.stream_index -
|
|
|
|
|
1];
|
|
|
|
|
transcodingParam.stream[2] =
|
|
|
|
|
ISACencUB_obj->bitstr_obj.stream[ISACencUB_obj->bitstr_obj.stream_index];
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Store LPC Gains before encoding them. */
|
|
|
|
|
for (k = 0; k < SUBFRAMES; k++) {
|
2011-07-07 08:21:25 +00:00
|
|
|
transcodingParam.loFiltGain[k] = lpcGains[k];
|
|
|
|
|
transcodingParam.hiFiltGain[k] = lpcGains[SUBFRAMES + k];
|
|
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Store the gains for multiple encoding. */
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
memcpy(ISACencUB_obj->SaveEnc_obj.lpcGain, lpcGains,
|
|
|
|
|
(SUBFRAMES << 1) * sizeof(double));
|
2011-07-07 08:21:25 +00:00
|
|
|
|
|
|
|
|
WebRtcIsac_EncodeLpcGainUb(lpcGains, &ISACencUB_obj->bitstr_obj,
|
|
|
|
|
ISACencUB_obj->SaveEnc_obj.lpcGainIndex);
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
WebRtcIsac_EncodeLpcGainUb(
|
|
|
|
|
&lpcGains[SUBFRAMES], &ISACencUB_obj->bitstr_obj,
|
|
|
|
|
&ISACencUB_obj->SaveEnc_obj.lpcGainIndex[SUBFRAMES]);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
|
|
|
|
/* Get the correct value for the payload limit and calculate the number of
|
|
|
|
|
bytes left for coding the spectrum. It is a 30ms frame
|
|
|
|
|
Subract 3 because termination process may add 3 bytes */
|
|
|
|
|
payloadLimitBytes = ISACencUB_obj->maxPayloadSizeBytes -
|
|
|
|
|
ISACencUB_obj->numBytesUsed - 3;
|
|
|
|
|
bytesLeftSpecCoding = payloadLimitBytes -
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
ISACencUB_obj->bitstr_obj.stream_index;
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
for (k = 0; k < (SUBFRAMES << 1); k++) {
|
|
|
|
|
percepFilterParams[k * (UB_LPC_ORDER + 1) + (UB_LPC_ORDER + 1)] =
|
2011-07-07 08:21:25 +00:00
|
|
|
lpcGains[k];
|
|
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* LPC filtering (using normalized lattice filter), */
|
|
|
|
|
/* first half-frame. */
|
2011-07-07 08:21:25 +00:00
|
|
|
WebRtcIsac_NormLatticeFilterMa(UB_LPC_ORDER,
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
ISACencUB_obj->maskfiltstr_obj.PreStateLoF,
|
|
|
|
|
ISACencUB_obj->maskfiltstr_obj.PreStateLoG,
|
|
|
|
|
&ISACencUB_obj->data_buffer_float[0],
|
|
|
|
|
&percepFilterParams[UB_LPC_ORDER + 1],
|
|
|
|
|
&LP_lookahead[0]);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Second half-frame filtering. */
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
WebRtcIsac_NormLatticeFilterMa(
|
|
|
|
|
UB_LPC_ORDER, ISACencUB_obj->maskfiltstr_obj.PreStateLoF,
|
|
|
|
|
ISACencUB_obj->maskfiltstr_obj.PreStateLoG,
|
|
|
|
|
&ISACencUB_obj->data_buffer_float[FRAMESAMPLES_HALF],
|
|
|
|
|
&percepFilterParams[(UB_LPC_ORDER + 1) + SUBFRAMES * (UB_LPC_ORDER + 1)],
|
|
|
|
|
&LP_lookahead[FRAMESAMPLES_HALF]);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
2015-06-16 00:02:34 +02:00
|
|
|
WebRtcIsac_Time2Spec(transform_tables,
|
|
|
|
|
&LP_lookahead[0], &LP_lookahead[FRAMESAMPLES_HALF],
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
fre, fim, &ISACencUB_obj->fftstr_obj);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Store FFT coefficients for multiple encoding. */
|
|
|
|
|
memcpy(ISACencUB_obj->SaveEnc_obj.realFFT, fre, sizeof(fre));
|
|
|
|
|
memcpy(ISACencUB_obj->SaveEnc_obj.imagFFT, fim, sizeof(fim));
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Prepare the audio buffer for the next packet
|
|
|
|
|
* move the last 3 ms to the beginning of the buffer. */
|
2011-07-07 08:21:25 +00:00
|
|
|
memcpy(ISACencUB_obj->data_buffer_float,
|
|
|
|
|
&ISACencUB_obj->data_buffer_float[FRAMESAMPLES],
|
|
|
|
|
LB_TOTAL_DELAY_SAMPLES * sizeof(float));
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* start writing with 3 ms delay to compensate for the delay
|
|
|
|
|
* of the lower-band. */
|
2011-07-07 08:21:25 +00:00
|
|
|
ISACencUB_obj->buffer_index = LB_TOTAL_DELAY_SAMPLES;
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Save the bit-stream object at this point for FEC. */
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
memcpy(&ISACencUB_obj->SaveEnc_obj.bitStreamObj, &ISACencUB_obj->bitstr_obj,
|
|
|
|
|
sizeof(Bitstr));
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Qantization and lossless coding */
|
|
|
|
|
/* Note that there is no pitch-gain for this band so kAveragePitchGain = 0
|
|
|
|
|
* is passed to the function. In fact, the function ignores the 3rd parameter
|
|
|
|
|
* for this band. */
|
|
|
|
|
err = WebRtcIsac_EncodeSpec(fre, fim, kAveragePitchGain, kIsacUpperBand16,
|
|
|
|
|
&ISACencUB_obj->bitstr_obj);
|
2011-07-07 08:21:25 +00:00
|
|
|
if ((err < 0) && (err != -ISAC_DISALLOWED_BITSTREAM_LENGTH)) {
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
if ((ISACencUB_obj->bitstr_obj.stream_index > payloadLimitBytes) ||
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
(err == -ISAC_DISALLOWED_BITSTREAM_LENGTH)) {
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
err = LimitPayloadUb(ISACencUB_obj, payloadLimitBytes, bytesLeftSpecCoding,
|
|
|
|
|
&transcodingParam, fre, fim, lpcGains,
|
|
|
|
|
kIsacUpperBand16, err);
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
if (err < 0) {
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
/* Complete arithmetic coding. */
|
2011-07-07 08:21:25 +00:00
|
|
|
return WebRtcIsac_EncTerminate(&ISACencUB_obj->bitstr_obj);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2015-06-16 00:02:34 +02:00
|
|
|
int WebRtcIsac_EncodeUb12(const TransformTables* transform_tables,
|
|
|
|
|
float* in, ISACUBEncStruct* ISACencUB_obj,
|
2013-04-09 00:28:06 +00:00
|
|
|
int32_t jitterInfo) {
|
2011-07-07 08:21:25 +00:00
|
|
|
int err;
|
|
|
|
|
int k;
|
|
|
|
|
|
|
|
|
|
double lpcVecs[UB_LPC_ORDER * UB_LPC_VEC_PER_FRAME];
|
|
|
|
|
|
|
|
|
|
double percepFilterParams[(1 + UB_LPC_ORDER) * SUBFRAMES];
|
|
|
|
|
float LP[FRAMESAMPLES_HALF];
|
|
|
|
|
float HP[FRAMESAMPLES_HALF];
|
|
|
|
|
|
|
|
|
|
double LP_lookahead[FRAMESAMPLES_HALF];
|
|
|
|
|
double HP_lookahead[FRAMESAMPLES_HALF];
|
|
|
|
|
double LPw[FRAMESAMPLES_HALF];
|
|
|
|
|
|
|
|
|
|
double HPw[FRAMESAMPLES_HALF];
|
2013-04-09 00:28:06 +00:00
|
|
|
int16_t fre[FRAMESAMPLES_HALF]; /* Q7 */
|
|
|
|
|
int16_t fim[FRAMESAMPLES_HALF]; /* Q7 */
|
2011-07-07 08:21:25 +00:00
|
|
|
|
|
|
|
|
int status = 0;
|
|
|
|
|
|
|
|
|
|
double varscale[1];
|
|
|
|
|
|
|
|
|
|
double corr[UB_LPC_GAIN_DIM][UB_LPC_ORDER + 1];
|
|
|
|
|
double lpcGains[SUBFRAMES];
|
|
|
|
|
transcode_obj transcodingParam;
|
2013-04-09 00:28:06 +00:00
|
|
|
uint16_t payloadLimitBytes;
|
2011-07-07 08:21:25 +00:00
|
|
|
double s2nr;
|
2013-04-09 00:28:06 +00:00
|
|
|
const int16_t kAveragePitchGain = 0.0;
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
double bytesLeftSpecCoding;
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Buffer speech samples (by 10ms packet) until the framelength is */
|
|
|
|
|
/* reached (30 ms). */
|
2011-07-07 08:21:25 +00:00
|
|
|
/********************************************************************/
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Fill the buffer with 10ms input data. */
|
|
|
|
|
memcpy(&ISACencUB_obj->data_buffer_float[ISACencUB_obj->buffer_index], in,
|
|
|
|
|
FRAMESAMPLES_10ms * sizeof(float));
|
2011-07-07 08:21:25 +00:00
|
|
|
|
|
|
|
|
/* if buffer-size is not equal to current frame-size then increase the
|
|
|
|
|
index and return. We do the encoding when we have enough audio. */
|
|
|
|
|
if (ISACencUB_obj->buffer_index + FRAMESAMPLES_10ms < FRAMESAMPLES) {
|
|
|
|
|
ISACencUB_obj->buffer_index += FRAMESAMPLES_10ms;
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* If buffer reached the right size, reset index and continue
|
2011-07-07 08:21:25 +00:00
|
|
|
with encoding the frame */
|
|
|
|
|
ISACencUB_obj->buffer_index = 0;
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* End of buffer function */
|
2011-07-07 08:21:25 +00:00
|
|
|
/**************************/
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Encoding */
|
2011-07-07 08:21:25 +00:00
|
|
|
/************/
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Reset bit-stream. */
|
|
|
|
|
WebRtcIsac_ResetBitstream(&(ISACencUB_obj->bitstr_obj));
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Encoding bandwidth information. */
|
2011-07-07 08:21:25 +00:00
|
|
|
WebRtcIsac_EncodeJitterInfo(jitterInfo, &ISACencUB_obj->bitstr_obj);
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
status = WebRtcIsac_EncodeBandwidth(isac12kHz, &ISACencUB_obj->bitstr_obj);
|
2011-07-07 08:21:25 +00:00
|
|
|
if (status < 0) {
|
|
|
|
|
return status;
|
|
|
|
|
}
|
|
|
|
|
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
s2nr = WebRtcIsac_GetSnr(ISACencUB_obj->bottleneck, FRAMESAMPLES);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Split signal in two bands. */
|
2011-07-07 08:21:25 +00:00
|
|
|
WebRtcIsac_SplitAndFilterFloat(ISACencUB_obj->data_buffer_float, HP, LP,
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
HP_lookahead, LP_lookahead,
|
|
|
|
|
&ISACencUB_obj->prefiltbankstr_obj);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Find coefficients for perceptual pre-filters. */
|
2011-07-07 08:21:25 +00:00
|
|
|
WebRtcIsac_GetLpcCoefUb(LP_lookahead, &ISACencUB_obj->maskfiltstr_obj,
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
lpcVecs, corr, varscale, isac12kHz);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Code LPC model and shape - gains not quantized yet. */
|
2011-07-07 08:21:25 +00:00
|
|
|
WebRtcIsac_EncodeLpcUB(lpcVecs, &ISACencUB_obj->bitstr_obj,
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
percepFilterParams, isac12kHz,
|
|
|
|
|
&ISACencUB_obj->SaveEnc_obj);
|
|
|
|
|
|
|
|
|
|
WebRtcIsac_GetLpcGain(s2nr, percepFilterParams, SUBFRAMES, lpcGains, corr,
|
|
|
|
|
varscale);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Store the state of arithmetic coder before coding LPC gains. */
|
2011-07-07 08:21:25 +00:00
|
|
|
transcodingParam.W_upper = ISACencUB_obj->bitstr_obj.W_upper;
|
|
|
|
|
transcodingParam.stream_index = ISACencUB_obj->bitstr_obj.stream_index;
|
|
|
|
|
transcodingParam.streamval = ISACencUB_obj->bitstr_obj.streamval;
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
transcodingParam.stream[0] =
|
|
|
|
|
ISACencUB_obj->bitstr_obj.stream[ISACencUB_obj->bitstr_obj.stream_index -
|
|
|
|
|
2];
|
|
|
|
|
transcodingParam.stream[1] =
|
|
|
|
|
ISACencUB_obj->bitstr_obj.stream[ISACencUB_obj->bitstr_obj.stream_index -
|
|
|
|
|
1];
|
|
|
|
|
transcodingParam.stream[2] =
|
|
|
|
|
ISACencUB_obj->bitstr_obj.stream[ISACencUB_obj->bitstr_obj.stream_index];
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Store LPC Gains before encoding them. */
|
|
|
|
|
for (k = 0; k < SUBFRAMES; k++) {
|
2011-07-07 08:21:25 +00:00
|
|
|
transcodingParam.loFiltGain[k] = lpcGains[k];
|
|
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Store the gains for multiple encoding. */
|
2011-07-07 08:21:25 +00:00
|
|
|
memcpy(ISACencUB_obj->SaveEnc_obj.lpcGain, lpcGains, SUBFRAMES *
|
|
|
|
|
sizeof(double));
|
|
|
|
|
|
|
|
|
|
WebRtcIsac_EncodeLpcGainUb(lpcGains, &ISACencUB_obj->bitstr_obj,
|
|
|
|
|
ISACencUB_obj->SaveEnc_obj.lpcGainIndex);
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
for (k = 0; k < SUBFRAMES; k++) {
|
|
|
|
|
percepFilterParams[k * (UB_LPC_ORDER + 1)] = lpcGains[k];
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* perceptual pre-filtering (using normalized lattice filter) */
|
|
|
|
|
/* low-band filtering */
|
|
|
|
|
WebRtcIsac_NormLatticeFilterMa(UB_LPC_ORDER,
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
ISACencUB_obj->maskfiltstr_obj.PreStateLoF,
|
|
|
|
|
ISACencUB_obj->maskfiltstr_obj.PreStateLoG, LP,
|
|
|
|
|
percepFilterParams, LPw);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
|
|
|
|
/* Get the correct value for the payload limit and calculate the number
|
|
|
|
|
of bytes left for coding the spectrum. It is a 30ms frame Subract 3
|
|
|
|
|
because termination process may add 3 bytes */
|
|
|
|
|
payloadLimitBytes = ISACencUB_obj->maxPayloadSizeBytes -
|
|
|
|
|
ISACencUB_obj->numBytesUsed - 3;
|
|
|
|
|
bytesLeftSpecCoding = payloadLimitBytes -
|
|
|
|
|
ISACencUB_obj->bitstr_obj.stream_index;
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
memset(HPw, 0, sizeof(HPw));
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Transform */
|
2015-06-16 00:02:34 +02:00
|
|
|
WebRtcIsac_Time2Spec(transform_tables,
|
|
|
|
|
LPw, HPw, fre, fim, &ISACencUB_obj->fftstr_obj);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Store FFT coefficients for multiple encoding. */
|
|
|
|
|
memcpy(ISACencUB_obj->SaveEnc_obj.realFFT, fre,
|
|
|
|
|
sizeof(ISACencUB_obj->SaveEnc_obj.realFFT));
|
|
|
|
|
memcpy(ISACencUB_obj->SaveEnc_obj.imagFFT, fim,
|
|
|
|
|
sizeof(ISACencUB_obj->SaveEnc_obj.imagFFT));
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Save the bit-stream object at this point for FEC. */
|
2011-07-07 08:21:25 +00:00
|
|
|
memcpy(&ISACencUB_obj->SaveEnc_obj.bitStreamObj,
|
|
|
|
|
&ISACencUB_obj->bitstr_obj, sizeof(Bitstr));
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Quantization and loss-less coding */
|
|
|
|
|
/* The 4th parameter to this function is pitch-gain, which is only used
|
|
|
|
|
* when encoding 0-8 kHz band, and irrelevant in this function, therefore,
|
|
|
|
|
* we insert zero here. */
|
|
|
|
|
err = WebRtcIsac_EncodeSpec(fre, fim, kAveragePitchGain, kIsacUpperBand12,
|
|
|
|
|
&ISACencUB_obj->bitstr_obj);
|
2011-07-07 08:21:25 +00:00
|
|
|
if ((err < 0) && (err != -ISAC_DISALLOWED_BITSTREAM_LENGTH)) {
|
|
|
|
|
/* There has been an error but it was not too large
|
|
|
|
|
payload (we can cure too large payload) */
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
if ((ISACencUB_obj->bitstr_obj.stream_index > payloadLimitBytes) ||
|
|
|
|
|
(err == -ISAC_DISALLOWED_BITSTREAM_LENGTH)) {
|
|
|
|
|
err = LimitPayloadUb(ISACencUB_obj, payloadLimitBytes, bytesLeftSpecCoding,
|
|
|
|
|
&transcodingParam, fre, fim, lpcGains,
|
|
|
|
|
kIsacUpperBand12, err);
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
if (err < 0) {
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
/* Complete arithmetic coding. */
|
2011-07-07 08:21:25 +00:00
|
|
|
return WebRtcIsac_EncTerminate(&ISACencUB_obj->bitstr_obj);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* This function is used to create a new bit-stream with new BWE.
|
2011-07-07 08:21:25 +00:00
|
|
|
The same data as previously encoded with the function WebRtcIsac_Encoder().
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
The data needed is taken from the structure, where it was stored
|
2011-07-07 08:21:25 +00:00
|
|
|
when calling the encoder. */
|
|
|
|
|
|
2014-12-17 15:23:29 +00:00
|
|
|
int WebRtcIsac_EncodeStoredDataLb(const IsacSaveEncoderData* ISACSavedEnc_obj,
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
Bitstr* ISACBitStr_obj, int BWnumber,
|
|
|
|
|
float scale) {
|
2011-07-07 08:21:25 +00:00
|
|
|
int ii;
|
|
|
|
|
int status;
|
|
|
|
|
int BWno = BWnumber;
|
|
|
|
|
|
2013-04-09 00:28:06 +00:00
|
|
|
const uint16_t* WebRtcIsac_kQPitchGainCdf_ptr[1];
|
|
|
|
|
const uint16_t** cdf;
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
double tmpLPCcoeffs_lo[(ORDERLO + 1)*SUBFRAMES * 2];
|
|
|
|
|
double tmpLPCcoeffs_hi[(ORDERHI + 1)*SUBFRAMES * 2];
|
|
|
|
|
int tmpLPCindex_g[12 * 2];
|
2013-04-09 00:28:06 +00:00
|
|
|
int16_t tmp_fre[FRAMESAMPLES], tmp_fim[FRAMESAMPLES];
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
const int kModel = 0;
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Sanity Check - possible values for BWnumber is 0 - 23. */
|
2011-07-07 08:21:25 +00:00
|
|
|
if ((BWnumber < 0) || (BWnumber > 23)) {
|
|
|
|
|
return -ISAC_RANGE_ERROR_BW_ESTIMATOR;
|
|
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Reset bit-stream. */
|
|
|
|
|
WebRtcIsac_ResetBitstream(ISACBitStr_obj);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Encode frame length */
|
2011-07-07 08:21:25 +00:00
|
|
|
status = WebRtcIsac_EncodeFrameLen(ISACSavedEnc_obj->framelength,
|
|
|
|
|
ISACBitStr_obj);
|
|
|
|
|
if (status < 0) {
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Wrong frame size. */
|
2011-07-07 08:21:25 +00:00
|
|
|
return status;
|
|
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Transcoding */
|
2011-07-07 08:21:25 +00:00
|
|
|
if ((scale > 0.0) && (scale < 1.0)) {
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Compensate LPC gain. */
|
2011-07-07 08:21:25 +00:00
|
|
|
for (ii = 0;
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
ii < ((ORDERLO + 1)* SUBFRAMES * (1 + ISACSavedEnc_obj->startIdx));
|
|
|
|
|
ii++) {
|
2011-07-07 08:21:25 +00:00
|
|
|
tmpLPCcoeffs_lo[ii] = scale * ISACSavedEnc_obj->LPCcoeffs_lo[ii];
|
|
|
|
|
}
|
|
|
|
|
for (ii = 0;
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
ii < ((ORDERHI + 1) * SUBFRAMES * (1 + ISACSavedEnc_obj->startIdx));
|
|
|
|
|
ii++) {
|
2011-07-07 08:21:25 +00:00
|
|
|
tmpLPCcoeffs_hi[ii] = scale * ISACSavedEnc_obj->LPCcoeffs_hi[ii];
|
|
|
|
|
}
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Scale DFT. */
|
2011-07-07 08:21:25 +00:00
|
|
|
for (ii = 0;
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
ii < (FRAMESAMPLES_HALF * (1 + ISACSavedEnc_obj->startIdx));
|
|
|
|
|
ii++) {
|
2013-04-09 00:28:06 +00:00
|
|
|
tmp_fre[ii] = (int16_t)((scale) * (float)ISACSavedEnc_obj->fre[ii]);
|
|
|
|
|
tmp_fim[ii] = (int16_t)((scale) * (float)ISACSavedEnc_obj->fim[ii]);
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
} else {
|
|
|
|
|
for (ii = 0;
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
ii < (KLT_ORDER_GAIN * (1 + ISACSavedEnc_obj->startIdx));
|
|
|
|
|
ii++) {
|
2011-07-07 08:21:25 +00:00
|
|
|
tmpLPCindex_g[ii] = ISACSavedEnc_obj->LPCindex_g[ii];
|
|
|
|
|
}
|
|
|
|
|
for (ii = 0;
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
ii < (FRAMESAMPLES_HALF * (1 + ISACSavedEnc_obj->startIdx));
|
|
|
|
|
ii++) {
|
2011-07-07 08:21:25 +00:00
|
|
|
tmp_fre[ii] = ISACSavedEnc_obj->fre[ii];
|
|
|
|
|
tmp_fim[ii] = ISACSavedEnc_obj->fim[ii];
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Encode bandwidth estimate. */
|
2011-07-07 08:21:25 +00:00
|
|
|
WebRtcIsac_EncodeReceiveBw(&BWno, ISACBitStr_obj);
|
|
|
|
|
|
|
|
|
|
/* Loop over number of 30 msec */
|
|
|
|
|
for (ii = 0; ii <= ISACSavedEnc_obj->startIdx; ii++) {
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Encode pitch gains. */
|
2011-07-07 08:21:25 +00:00
|
|
|
*WebRtcIsac_kQPitchGainCdf_ptr = WebRtcIsac_kQPitchGainCdf;
|
|
|
|
|
WebRtcIsac_EncHistMulti(ISACBitStr_obj,
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
&ISACSavedEnc_obj->pitchGain_index[ii],
|
|
|
|
|
WebRtcIsac_kQPitchGainCdf_ptr, 1);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Entropy coding of quantization pitch lags */
|
|
|
|
|
/* Voicing classification. */
|
2011-07-07 08:21:25 +00:00
|
|
|
if (ISACSavedEnc_obj->meanGain[ii] < 0.2) {
|
|
|
|
|
cdf = WebRtcIsac_kQPitchLagCdfPtrLo;
|
|
|
|
|
} else if (ISACSavedEnc_obj->meanGain[ii] < 0.4) {
|
|
|
|
|
cdf = WebRtcIsac_kQPitchLagCdfPtrMid;
|
|
|
|
|
} else {
|
|
|
|
|
cdf = WebRtcIsac_kQPitchLagCdfPtrHi;
|
|
|
|
|
}
|
|
|
|
|
WebRtcIsac_EncHistMulti(ISACBitStr_obj,
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
&ISACSavedEnc_obj->pitchIndex[PITCH_SUBFRAMES * ii],
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
cdf, PITCH_SUBFRAMES);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
|
|
|
|
/* LPC */
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Only one model exists. The entropy coding is done only for backward
|
|
|
|
|
* compatibility. */
|
|
|
|
|
WebRtcIsac_EncHistMulti(ISACBitStr_obj, &kModel,
|
|
|
|
|
WebRtcIsac_kQKltModelCdfPtr, 1);
|
|
|
|
|
/* Entropy coding of quantization indices - LPC shape only. */
|
2011-07-07 08:21:25 +00:00
|
|
|
WebRtcIsac_EncHistMulti(ISACBitStr_obj,
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
&ISACSavedEnc_obj->LPCindex_s[KLT_ORDER_SHAPE * ii],
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
WebRtcIsac_kQKltCdfPtrShape,
|
2011-07-07 08:21:25 +00:00
|
|
|
KLT_ORDER_SHAPE);
|
|
|
|
|
|
|
|
|
|
/* If transcoding, get new LPC gain indices */
|
|
|
|
|
if (scale < 1.0) {
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
WebRtcIsac_TranscodeLPCCoef(
|
|
|
|
|
&tmpLPCcoeffs_lo[(ORDERLO + 1) * SUBFRAMES * ii],
|
|
|
|
|
&tmpLPCcoeffs_hi[(ORDERHI + 1)*SUBFRAMES * ii],
|
|
|
|
|
&tmpLPCindex_g[KLT_ORDER_GAIN * ii]);
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Entropy coding of quantization indices - LPC gain. */
|
|
|
|
|
WebRtcIsac_EncHistMulti(ISACBitStr_obj, &tmpLPCindex_g[KLT_ORDER_GAIN * ii],
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
WebRtcIsac_kQKltCdfPtrGain, KLT_ORDER_GAIN);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Quantization and loss-less coding. */
|
|
|
|
|
status = WebRtcIsac_EncodeSpec(&tmp_fre[ii * FRAMESAMPLES_HALF],
|
|
|
|
|
&tmp_fim[ii * FRAMESAMPLES_HALF],
|
|
|
|
|
ISACSavedEnc_obj->AvgPitchGain[ii],
|
|
|
|
|
kIsacLowerBand, ISACBitStr_obj);
|
2011-07-07 08:21:25 +00:00
|
|
|
if (status < 0) {
|
|
|
|
|
return status;
|
|
|
|
|
}
|
|
|
|
|
}
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Complete arithmetic coding. */
|
2011-07-07 08:21:25 +00:00
|
|
|
return WebRtcIsac_EncTerminate(ISACBitStr_obj);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
int WebRtcIsac_EncodeStoredDataUb(
|
2011-07-07 08:21:25 +00:00
|
|
|
const ISACUBSaveEncDataStruct* ISACSavedEnc_obj,
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
Bitstr* bitStream,
|
2013-04-09 00:28:06 +00:00
|
|
|
int32_t jitterInfo,
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
float scale,
|
|
|
|
|
enum ISACBandwidth bandwidth) {
|
2011-07-07 08:21:25 +00:00
|
|
|
int n;
|
|
|
|
|
int err;
|
|
|
|
|
double lpcGain[SUBFRAMES];
|
2013-04-09 00:28:06 +00:00
|
|
|
int16_t realFFT[FRAMESAMPLES_HALF];
|
|
|
|
|
int16_t imagFFT[FRAMESAMPLES_HALF];
|
|
|
|
|
const uint16_t** shape_cdf;
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
int shape_len;
|
2013-04-09 00:28:06 +00:00
|
|
|
const int16_t kAveragePitchGain = 0.0;
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
enum ISACBand band;
|
|
|
|
|
/* Reset bitstream. */
|
|
|
|
|
WebRtcIsac_ResetBitstream(bitStream);
|
|
|
|
|
|
|
|
|
|
/* Encode jitter index. */
|
2011-07-07 08:21:25 +00:00
|
|
|
WebRtcIsac_EncodeJitterInfo(jitterInfo, bitStream);
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
err = WebRtcIsac_EncodeBandwidth(bandwidth, bitStream);
|
|
|
|
|
if (err < 0) {
|
2011-07-07 08:21:25 +00:00
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Encode LPC-shape. */
|
|
|
|
|
if (bandwidth == isac12kHz) {
|
|
|
|
|
shape_cdf = WebRtcIsac_kLpcShapeCdfMatUb12;
|
|
|
|
|
shape_len = UB_LPC_ORDER * UB_LPC_VEC_PER_FRAME;
|
|
|
|
|
band = kIsacUpperBand12;
|
|
|
|
|
} else {
|
|
|
|
|
shape_cdf = WebRtcIsac_kLpcShapeCdfMatUb16;
|
|
|
|
|
shape_len = UB_LPC_ORDER * UB16_LPC_VEC_PER_FRAME;
|
|
|
|
|
band = kIsacUpperBand16;
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
WebRtcIsac_EncHistMulti(bitStream, ISACSavedEnc_obj->indexLPCShape,
|
|
|
|
|
shape_cdf, shape_len);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
if ((scale <= 0.0) || (scale >= 1.0)) {
|
|
|
|
|
/* We only consider scales between zero and one. */
|
2011-07-07 08:21:25 +00:00
|
|
|
WebRtcIsac_EncHistMulti(bitStream, ISACSavedEnc_obj->lpcGainIndex,
|
|
|
|
|
WebRtcIsac_kLpcGainCdfMat, UB_LPC_GAIN_DIM);
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
if (bandwidth == isac16kHz) {
|
|
|
|
|
/* Store gain indices of the second half. */
|
|
|
|
|
WebRtcIsac_EncHistMulti(bitStream,
|
|
|
|
|
&ISACSavedEnc_obj->lpcGainIndex[SUBFRAMES],
|
|
|
|
|
WebRtcIsac_kLpcGainCdfMat, UB_LPC_GAIN_DIM);
|
|
|
|
|
}
|
|
|
|
|
/* Store FFT coefficients. */
|
|
|
|
|
err = WebRtcIsac_EncodeSpec(ISACSavedEnc_obj->realFFT,
|
|
|
|
|
ISACSavedEnc_obj->imagFFT, kAveragePitchGain,
|
|
|
|
|
band, bitStream);
|
|
|
|
|
} else {
|
|
|
|
|
/* Scale LPC gain and FFT coefficients. */
|
|
|
|
|
for (n = 0; n < SUBFRAMES; n++) {
|
2011-07-07 08:21:25 +00:00
|
|
|
lpcGain[n] = scale * ISACSavedEnc_obj->lpcGain[n];
|
|
|
|
|
}
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Store LPC gains. */
|
2011-07-07 08:21:25 +00:00
|
|
|
WebRtcIsac_StoreLpcGainUb(lpcGain, bitStream);
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
if (bandwidth == isac16kHz) {
|
|
|
|
|
/* Scale and code the gains of the second half of the frame, if 16kHz. */
|
|
|
|
|
for (n = 0; n < SUBFRAMES; n++) {
|
|
|
|
|
lpcGain[n] = scale * ISACSavedEnc_obj->lpcGain[n + SUBFRAMES];
|
|
|
|
|
}
|
|
|
|
|
WebRtcIsac_StoreLpcGainUb(lpcGain, bitStream);
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
|
|
|
|
|
for (n = 0; n < FRAMESAMPLES_HALF; n++) {
|
2013-04-09 00:28:06 +00:00
|
|
|
realFFT[n] = (int16_t)(scale * (float)ISACSavedEnc_obj->realFFT[n] +
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
0.5f);
|
2013-04-09 00:28:06 +00:00
|
|
|
imagFFT[n] = (int16_t)(scale * (float)ISACSavedEnc_obj->imagFFT[n] +
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
0.5f);
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Store FFT coefficients. */
|
|
|
|
|
err = WebRtcIsac_EncodeSpec(realFFT, imagFFT, kAveragePitchGain,
|
|
|
|
|
band, bitStream);
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
if (err < 0) {
|
|
|
|
|
/* Error happened while encoding FFT coefficients. */
|
2011-07-07 08:21:25 +00:00
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Complete arithmetic coding. */
|
2011-07-07 08:21:25 +00:00
|
|
|
return WebRtcIsac_EncTerminate(bitStream);
|
|
|
|
|
}
|
|
|
|
|
|
2013-04-09 00:28:06 +00:00
|
|
|
int16_t WebRtcIsac_GetRedPayloadUb(
|
2011-07-07 08:21:25 +00:00
|
|
|
const ISACUBSaveEncDataStruct* ISACSavedEncObj,
|
|
|
|
|
Bitstr* bitStreamObj,
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
enum ISACBandwidth bandwidth) {
|
2011-07-07 08:21:25 +00:00
|
|
|
int n;
|
2013-04-09 00:28:06 +00:00
|
|
|
int16_t status;
|
|
|
|
|
int16_t realFFT[FRAMESAMPLES_HALF];
|
|
|
|
|
int16_t imagFFT[FRAMESAMPLES_HALF];
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
enum ISACBand band;
|
2013-04-09 00:28:06 +00:00
|
|
|
const int16_t kAveragePitchGain = 0.0;
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Store bit-stream object. */
|
2011-07-07 08:21:25 +00:00
|
|
|
memcpy(bitStreamObj, &ISACSavedEncObj->bitStreamObj, sizeof(Bitstr));
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
/* Scale FFT coefficients. */
|
|
|
|
|
for (n = 0; n < FRAMESAMPLES_HALF; n++) {
|
2013-04-09 00:28:06 +00:00
|
|
|
realFFT[n] = (int16_t)((float)ISACSavedEncObj->realFFT[n] *
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
RCU_TRANSCODING_SCALE_UB + 0.5);
|
2013-04-09 00:28:06 +00:00
|
|
|
imagFFT[n] = (int16_t)((float)ISACSavedEncObj->imagFFT[n] *
|
Hi Tina,
I have uploaded this patch for your review. I have done an extensive test to be sure that removing of tables does not create any problem.
The test file, is called test_lpc.c which requires a hack to standard iSAC. The test computes LPC coefficients, then encodes and decodes with old and new (size-reduced) tables. It compares the results is all steps. I have ran the test over large set of files, more then 51 hours of audio, and there was no error.
I tried to do no formatting so the review to be easier, but I know it can be a tricky CL. Hopefully, the test file helps you to be more confident on the CL.
Thanks,... Turaj
In this change list the LPC tables associated with mode 1 & 2 are remoded, and necessary cahnges are made to other files.
The only model allowed is model number 0. Therefore, this CL breaks compatibility with iSAC released prior to 2.4.3. To avoid changing the bit-stream, we still keep the model number in the bit-stream.
entropy_coding.c is cleaned up, especially encoding of LAR had KLT transform of LPC gains which are removed now.
Review URL: https://webrtc-codereview.appspot.com/548004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2186 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-05-07 20:36:22 +00:00
|
|
|
RCU_TRANSCODING_SCALE_UB + 0.5);
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
band = (bandwidth == isac12kHz) ? kIsacUpperBand12 : kIsacUpperBand16;
|
|
|
|
|
status = WebRtcIsac_EncodeSpec(realFFT, imagFFT, kAveragePitchGain, band,
|
|
|
|
|
bitStreamObj);
|
|
|
|
|
if (status < 0) {
|
2011-07-07 08:21:25 +00:00
|
|
|
return status;
|
This CL contains some refactoring. Spectrum coding is main place that is affected. Therefore, I have bit-exactness test, test_spectrum_
coding.c, to be sure about the changelist. You can go through the test to be sure the changes are tested. However, I don't intend to commi
t the test, as it would be a source of confusion and requires hack to iSAC to be able to run the test. It is basically a one-time test.
The part which not covered in this test is where we limit payload for super-wideband bit-stream. I'll add a test for that as well.
I kept format changes at minimum in all files except isac.c, which was in bad shape, and coding changes were minimum. I'm planning to uplo
ad following patches to this CL where I try to address formatting issues. But I don't intend to change variable names, for the moment.
The refactoring is not yet finished, so you would find part of the code which could be cleaned up, especially KLT transforms in entropy_co
ding.c
Review URL: https://webrtc-codereview.appspot.com/580004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2359 4adac7df-926f-26a2-2b94-8c16560cd09d
2012-06-04 20:04:58 +00:00
|
|
|
} else {
|
|
|
|
|
/* Terminate entropy coding */
|
2011-07-07 08:21:25 +00:00
|
|
|
return WebRtcIsac_EncTerminate(bitStreamObj);
|
|
|
|
|
}
|
|
|
|
|
}
|