2011-07-07 08:21:25 +00:00
|
|
|
/*
|
2012-01-30 20:51:15 +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.
|
|
|
|
|
*/
|
|
|
|
|
|
2013-03-01 16:36:19 +00:00
|
|
|
#include "webrtc/modules/audio_processing/audio_processing_impl.h"
|
2011-07-07 08:21:25 +00:00
|
|
|
|
2015-07-23 11:41:39 -07:00
|
|
|
#include <algorithm>
|
2011-07-07 08:21:25 +00:00
|
|
|
|
2015-06-29 14:57:29 +02:00
|
|
|
#include "webrtc/base/checks.h"
|
2014-10-10 08:36:56 +00:00
|
|
|
#include "webrtc/base/platform_file.h"
|
2015-12-17 06:42:29 -08:00
|
|
|
#include "webrtc/base/trace_event.h"
|
2015-08-14 10:35:55 -07:00
|
|
|
#include "webrtc/common_audio/audio_converter.h"
|
2015-03-25 16:37:27 -07:00
|
|
|
#include "webrtc/common_audio/channel_buffer.h"
|
2015-08-14 10:35:55 -07:00
|
|
|
#include "webrtc/common_audio/include/audio_util.h"
|
2014-01-07 17:45:09 +00:00
|
|
|
#include "webrtc/common_audio/signal_processing/include/signal_processing_library.h"
|
2015-06-29 14:57:29 +02:00
|
|
|
#include "webrtc/modules/audio_processing/aec/aec_core.h"
|
2014-12-15 09:41:24 +00:00
|
|
|
#include "webrtc/modules/audio_processing/agc/agc_manager_direct.h"
|
2013-03-01 16:36:19 +00:00
|
|
|
#include "webrtc/modules/audio_processing/audio_buffer.h"
|
2015-03-13 00:13:32 +00:00
|
|
|
#include "webrtc/modules/audio_processing/beamformer/nonlinear_beamformer.h"
|
2014-04-22 21:00:04 +00:00
|
|
|
#include "webrtc/modules/audio_processing/common.h"
|
2014-02-27 22:23:17 +00:00
|
|
|
#include "webrtc/modules/audio_processing/echo_cancellation_impl.h"
|
2013-03-01 16:36:19 +00:00
|
|
|
#include "webrtc/modules/audio_processing/echo_control_mobile_impl.h"
|
2016-02-13 16:40:47 -08:00
|
|
|
#include "webrtc/modules/audio_processing/gain_control_for_experimental_agc.h"
|
2013-03-01 16:36:19 +00:00
|
|
|
#include "webrtc/modules/audio_processing/gain_control_impl.h"
|
2016-08-26 07:16:04 -07:00
|
|
|
#if WEBRTC_INTELLIGIBILITY_ENHANCER
|
2015-08-14 10:35:55 -07:00
|
|
|
#include "webrtc/modules/audio_processing/intelligibility/intelligibility_enhancer.h"
|
2016-08-26 07:16:04 -07:00
|
|
|
#endif
|
2016-06-29 15:26:12 -07:00
|
|
|
#include "webrtc/modules/audio_processing/level_controller/level_controller.h"
|
2013-03-01 16:36:19 +00:00
|
|
|
#include "webrtc/modules/audio_processing/level_estimator_impl.h"
|
2016-11-22 07:24:52 -08:00
|
|
|
#include "webrtc/modules/audio_processing/low_cut_filter.h"
|
2013-03-01 16:36:19 +00:00
|
|
|
#include "webrtc/modules/audio_processing/noise_suppression_impl.h"
|
2016-10-28 05:39:16 -07:00
|
|
|
#include "webrtc/modules/audio_processing/residual_echo_detector.h"
|
2014-12-19 19:57:34 +00:00
|
|
|
#include "webrtc/modules/audio_processing/transient/transient_suppressor.h"
|
2013-03-01 16:36:19 +00:00
|
|
|
#include "webrtc/modules/audio_processing/voice_detection_impl.h"
|
2015-11-04 08:31:52 +01:00
|
|
|
#include "webrtc/modules/include/module_common_types.h"
|
2015-10-28 18:17:40 +01:00
|
|
|
#include "webrtc/system_wrappers/include/file_wrapper.h"
|
|
|
|
|
#include "webrtc/system_wrappers/include/logging.h"
|
|
|
|
|
#include "webrtc/system_wrappers/include/metrics.h"
|
2011-12-03 00:03:31 +00:00
|
|
|
|
|
|
|
|
#ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP
|
|
|
|
|
// Files generated at build-time by the protobuf compiler.
|
2012-03-16 21:36:00 +00:00
|
|
|
#ifdef WEBRTC_ANDROID_PLATFORM_BUILD
|
2012-10-22 21:21:52 +00:00
|
|
|
#include "external/webrtc/webrtc/modules/audio_processing/debug.pb.h"
|
2011-08-03 23:34:31 +00:00
|
|
|
#else
|
2016-02-09 08:13:06 -08:00
|
|
|
#include "webrtc/modules/audio_processing/debug.pb.h"
|
2011-08-03 23:34:31 +00:00
|
|
|
#endif
|
2011-12-03 00:03:31 +00:00
|
|
|
#endif // WEBRTC_AUDIOPROC_DEBUG_DUMP
|
2011-07-07 08:21:25 +00:00
|
|
|
|
2016-08-26 07:16:04 -07:00
|
|
|
// Check to verify that the define for the intelligibility enhancer is properly
|
|
|
|
|
// set.
|
|
|
|
|
#if !defined(WEBRTC_INTELLIGIBILITY_ENHANCER) || \
|
|
|
|
|
(WEBRTC_INTELLIGIBILITY_ENHANCER != 0 && \
|
|
|
|
|
WEBRTC_INTELLIGIBILITY_ENHANCER != 1)
|
|
|
|
|
#error "Set WEBRTC_INTELLIGIBILITY_ENHANCER to either 0 or 1"
|
|
|
|
|
#endif
|
|
|
|
|
|
2015-07-23 11:41:39 -07:00
|
|
|
#define RETURN_ON_ERR(expr) \
|
|
|
|
|
do { \
|
|
|
|
|
int err = (expr); \
|
|
|
|
|
if (err != kNoError) { \
|
|
|
|
|
return err; \
|
|
|
|
|
} \
|
2014-01-07 17:45:09 +00:00
|
|
|
} while (0)
|
|
|
|
|
|
2011-07-07 08:21:25 +00:00
|
|
|
namespace webrtc {
|
2016-03-16 18:26:35 -07:00
|
|
|
|
2016-09-13 07:49:33 -07:00
|
|
|
constexpr int AudioProcessing::kNativeSampleRatesHz[];
|
2016-03-16 18:26:35 -07:00
|
|
|
|
2015-07-23 11:41:39 -07:00
|
|
|
namespace {
|
|
|
|
|
|
|
|
|
|
static bool LayoutHasKeyboard(AudioProcessing::ChannelLayout layout) {
|
|
|
|
|
switch (layout) {
|
|
|
|
|
case AudioProcessing::kMono:
|
|
|
|
|
case AudioProcessing::kStereo:
|
|
|
|
|
return false;
|
|
|
|
|
case AudioProcessing::kMonoAndKeyboard:
|
|
|
|
|
case AudioProcessing::kStereoAndKeyboard:
|
|
|
|
|
return true;
|
|
|
|
|
}
|
|
|
|
|
|
2016-09-14 05:23:22 -07:00
|
|
|
RTC_NOTREACHED();
|
2015-07-23 11:41:39 -07:00
|
|
|
return false;
|
|
|
|
|
}
|
2016-03-16 18:26:35 -07:00
|
|
|
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
bool SampleRateSupportsMultiBand(int sample_rate_hz) {
|
2016-03-16 18:26:35 -07:00
|
|
|
return sample_rate_hz == AudioProcessing::kSampleRate32kHz ||
|
|
|
|
|
sample_rate_hz == AudioProcessing::kSampleRate48kHz;
|
|
|
|
|
}
|
|
|
|
|
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
int FindNativeProcessRateToUse(int minimum_rate, bool band_splitting_required) {
|
|
|
|
|
#ifdef WEBRTC_ARCH_ARM_FAMILY
|
2016-09-13 07:49:33 -07:00
|
|
|
constexpr int kMaxSplittingNativeProcessRate =
|
|
|
|
|
AudioProcessing::kSampleRate32kHz;
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
#else
|
2016-09-13 07:49:33 -07:00
|
|
|
constexpr int kMaxSplittingNativeProcessRate =
|
|
|
|
|
AudioProcessing::kSampleRate48kHz;
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
#endif
|
2016-09-13 07:49:33 -07:00
|
|
|
static_assert(
|
|
|
|
|
kMaxSplittingNativeProcessRate <= AudioProcessing::kMaxNativeSampleRateHz,
|
|
|
|
|
"");
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
const int uppermost_native_rate = band_splitting_required
|
|
|
|
|
? kMaxSplittingNativeProcessRate
|
|
|
|
|
: AudioProcessing::kSampleRate48kHz;
|
|
|
|
|
|
|
|
|
|
for (auto rate : AudioProcessing::kNativeSampleRatesHz) {
|
|
|
|
|
if (rate >= uppermost_native_rate) {
|
|
|
|
|
return uppermost_native_rate;
|
|
|
|
|
}
|
|
|
|
|
if (rate >= minimum_rate) {
|
2016-03-16 18:26:35 -07:00
|
|
|
return rate;
|
|
|
|
|
}
|
|
|
|
|
}
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
RTC_NOTREACHED();
|
|
|
|
|
return uppermost_native_rate;
|
2016-03-16 18:26:35 -07:00
|
|
|
}
|
|
|
|
|
|
2016-10-22 05:04:30 -07:00
|
|
|
// Maximum length that a frame of samples can have.
|
|
|
|
|
static const size_t kMaxAllowedValuesOfSamplesPerFrame = 160;
|
|
|
|
|
// Maximum number of frames to buffer in the render queue.
|
|
|
|
|
// TODO(peah): Decrease this once we properly handle hugely unbalanced
|
|
|
|
|
// reverse and forward call numbers.
|
|
|
|
|
static const size_t kMaxNumFramesToBuffer = 100;
|
|
|
|
|
|
2016-11-22 07:24:52 -08:00
|
|
|
class HighPassFilterImpl : public HighPassFilter {
|
|
|
|
|
public:
|
|
|
|
|
explicit HighPassFilterImpl(AudioProcessingImpl* apm) : apm_(apm) {}
|
|
|
|
|
~HighPassFilterImpl() override = default;
|
|
|
|
|
|
|
|
|
|
// HighPassFilter implementation.
|
|
|
|
|
int Enable(bool enable) override {
|
|
|
|
|
apm_->MutateConfig([enable](AudioProcessing::Config* config) {
|
|
|
|
|
config->high_pass_filter.enabled = enable;
|
|
|
|
|
});
|
|
|
|
|
|
|
|
|
|
return AudioProcessing::kNoError;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
bool is_enabled() const override {
|
|
|
|
|
return apm_->GetConfig().high_pass_filter.enabled;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
private:
|
|
|
|
|
AudioProcessingImpl* apm_;
|
|
|
|
|
RTC_DISALLOW_IMPLICIT_CONSTRUCTORS(HighPassFilterImpl);
|
|
|
|
|
};
|
|
|
|
|
|
2015-07-23 11:41:39 -07:00
|
|
|
} // namespace
|
2014-01-07 17:45:09 +00:00
|
|
|
|
|
|
|
|
// Throughout webrtc, it's assumed that success is represented by zero.
|
2015-01-14 10:51:54 +00:00
|
|
|
static_assert(AudioProcessing::kNoError == 0, "kNoError must be zero");
|
2014-01-07 17:45:09 +00:00
|
|
|
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
AudioProcessingImpl::ApmSubmoduleStates::ApmSubmoduleStates() {}
|
|
|
|
|
|
|
|
|
|
bool AudioProcessingImpl::ApmSubmoduleStates::Update(
|
2016-11-22 07:24:52 -08:00
|
|
|
bool low_cut_filter_enabled,
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
bool echo_canceller_enabled,
|
|
|
|
|
bool mobile_echo_controller_enabled,
|
2016-10-28 05:39:16 -07:00
|
|
|
bool residual_echo_detector_enabled,
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
bool noise_suppressor_enabled,
|
|
|
|
|
bool intelligibility_enhancer_enabled,
|
|
|
|
|
bool beamformer_enabled,
|
|
|
|
|
bool adaptive_gain_controller_enabled,
|
|
|
|
|
bool level_controller_enabled,
|
|
|
|
|
bool voice_activity_detector_enabled,
|
|
|
|
|
bool level_estimator_enabled,
|
|
|
|
|
bool transient_suppressor_enabled) {
|
|
|
|
|
bool changed = false;
|
2016-11-22 07:24:52 -08:00
|
|
|
changed |= (low_cut_filter_enabled != low_cut_filter_enabled_);
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
changed |= (echo_canceller_enabled != echo_canceller_enabled_);
|
|
|
|
|
changed |=
|
|
|
|
|
(mobile_echo_controller_enabled != mobile_echo_controller_enabled_);
|
2016-10-28 05:39:16 -07:00
|
|
|
changed |=
|
|
|
|
|
(residual_echo_detector_enabled != residual_echo_detector_enabled_);
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
changed |= (noise_suppressor_enabled != noise_suppressor_enabled_);
|
|
|
|
|
changed |=
|
|
|
|
|
(intelligibility_enhancer_enabled != intelligibility_enhancer_enabled_);
|
|
|
|
|
changed |= (beamformer_enabled != beamformer_enabled_);
|
|
|
|
|
changed |=
|
|
|
|
|
(adaptive_gain_controller_enabled != adaptive_gain_controller_enabled_);
|
|
|
|
|
changed |= (level_controller_enabled != level_controller_enabled_);
|
|
|
|
|
changed |= (level_estimator_enabled != level_estimator_enabled_);
|
|
|
|
|
changed |=
|
|
|
|
|
(voice_activity_detector_enabled != voice_activity_detector_enabled_);
|
|
|
|
|
changed |= (transient_suppressor_enabled != transient_suppressor_enabled_);
|
|
|
|
|
if (changed) {
|
2016-11-22 07:24:52 -08:00
|
|
|
low_cut_filter_enabled_ = low_cut_filter_enabled;
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
echo_canceller_enabled_ = echo_canceller_enabled;
|
|
|
|
|
mobile_echo_controller_enabled_ = mobile_echo_controller_enabled;
|
2016-10-28 05:39:16 -07:00
|
|
|
residual_echo_detector_enabled_ = residual_echo_detector_enabled;
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
noise_suppressor_enabled_ = noise_suppressor_enabled;
|
|
|
|
|
intelligibility_enhancer_enabled_ = intelligibility_enhancer_enabled;
|
|
|
|
|
beamformer_enabled_ = beamformer_enabled;
|
|
|
|
|
adaptive_gain_controller_enabled_ = adaptive_gain_controller_enabled;
|
|
|
|
|
level_controller_enabled_ = level_controller_enabled;
|
|
|
|
|
level_estimator_enabled_ = level_estimator_enabled;
|
|
|
|
|
voice_activity_detector_enabled_ = voice_activity_detector_enabled;
|
|
|
|
|
transient_suppressor_enabled_ = transient_suppressor_enabled;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
changed |= first_update_;
|
|
|
|
|
first_update_ = false;
|
|
|
|
|
return changed;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
bool AudioProcessingImpl::ApmSubmoduleStates::CaptureMultiBandSubModulesActive()
|
|
|
|
|
const {
|
|
|
|
|
#if WEBRTC_INTELLIGIBILITY_ENHANCER
|
|
|
|
|
return CaptureMultiBandProcessingActive() ||
|
2016-11-15 05:24:35 -08:00
|
|
|
intelligibility_enhancer_enabled_ ||
|
|
|
|
|
voice_activity_detector_enabled_ || residual_echo_detector_enabled_;
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
#else
|
2016-11-15 05:24:35 -08:00
|
|
|
return CaptureMultiBandProcessingActive() ||
|
|
|
|
|
voice_activity_detector_enabled_ || residual_echo_detector_enabled_;
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
#endif
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
bool AudioProcessingImpl::ApmSubmoduleStates::CaptureMultiBandProcessingActive()
|
|
|
|
|
const {
|
2016-11-22 07:24:52 -08:00
|
|
|
return low_cut_filter_enabled_ || echo_canceller_enabled_ ||
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
mobile_echo_controller_enabled_ || noise_suppressor_enabled_ ||
|
|
|
|
|
beamformer_enabled_ || adaptive_gain_controller_enabled_;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
bool AudioProcessingImpl::ApmSubmoduleStates::RenderMultiBandSubModulesActive()
|
|
|
|
|
const {
|
|
|
|
|
return RenderMultiBandProcessingActive() || echo_canceller_enabled_ ||
|
2016-11-15 05:24:35 -08:00
|
|
|
mobile_echo_controller_enabled_ || adaptive_gain_controller_enabled_ ||
|
|
|
|
|
residual_echo_detector_enabled_;
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
bool AudioProcessingImpl::ApmSubmoduleStates::RenderMultiBandProcessingActive()
|
|
|
|
|
const {
|
|
|
|
|
#if WEBRTC_INTELLIGIBILITY_ENHANCER
|
|
|
|
|
return intelligibility_enhancer_enabled_;
|
|
|
|
|
#else
|
|
|
|
|
return false;
|
|
|
|
|
#endif
|
|
|
|
|
}
|
|
|
|
|
|
2015-12-08 13:22:33 -08:00
|
|
|
struct AudioProcessingImpl::ApmPublicSubmodules {
|
2016-03-10 21:09:04 -08:00
|
|
|
ApmPublicSubmodules() {}
|
2015-12-08 13:22:33 -08:00
|
|
|
// Accessed externally of APM without any lock acquired.
|
2016-03-05 03:01:14 -08:00
|
|
|
std::unique_ptr<EchoCancellationImpl> echo_cancellation;
|
2016-03-10 12:54:25 -08:00
|
|
|
std::unique_ptr<EchoControlMobileImpl> echo_control_mobile;
|
2016-03-10 21:09:04 -08:00
|
|
|
std::unique_ptr<GainControlImpl> gain_control;
|
2016-02-19 07:04:49 -08:00
|
|
|
std::unique_ptr<LevelEstimatorImpl> level_estimator;
|
|
|
|
|
std::unique_ptr<NoiseSuppressionImpl> noise_suppression;
|
|
|
|
|
std::unique_ptr<VoiceDetectionImpl> voice_detection;
|
|
|
|
|
std::unique_ptr<GainControlForExperimentalAgc>
|
2016-02-13 16:40:47 -08:00
|
|
|
gain_control_for_experimental_agc;
|
2015-12-08 13:22:33 -08:00
|
|
|
|
|
|
|
|
// Accessed internally from both render and capture.
|
2016-02-19 07:04:49 -08:00
|
|
|
std::unique_ptr<TransientSuppressor> transient_suppressor;
|
2016-08-26 07:16:04 -07:00
|
|
|
#if WEBRTC_INTELLIGIBILITY_ENHANCER
|
2016-02-19 07:04:49 -08:00
|
|
|
std::unique_ptr<IntelligibilityEnhancer> intelligibility_enhancer;
|
2016-08-26 07:16:04 -07:00
|
|
|
#endif
|
2015-12-08 13:22:33 -08:00
|
|
|
};
|
|
|
|
|
|
|
|
|
|
struct AudioProcessingImpl::ApmPrivateSubmodules {
|
2016-07-01 17:19:09 -07:00
|
|
|
explicit ApmPrivateSubmodules(NonlinearBeamformer* beamformer)
|
2015-12-08 13:22:33 -08:00
|
|
|
: beamformer(beamformer) {}
|
|
|
|
|
// Accessed internally from capture or during initialization
|
2016-07-01 17:19:09 -07:00
|
|
|
std::unique_ptr<NonlinearBeamformer> beamformer;
|
2016-02-19 07:04:49 -08:00
|
|
|
std::unique_ptr<AgcManagerDirect> agc_manager;
|
2016-11-22 07:24:52 -08:00
|
|
|
std::unique_ptr<LowCutFilter> low_cut_filter;
|
2016-06-29 15:26:12 -07:00
|
|
|
std::unique_ptr<LevelController> level_controller;
|
2016-10-28 05:39:16 -07:00
|
|
|
std::unique_ptr<ResidualEchoDetector> residual_echo_detector;
|
2015-12-08 13:22:33 -08:00
|
|
|
};
|
|
|
|
|
|
2014-01-25 02:09:06 +00:00
|
|
|
AudioProcessing* AudioProcessing::Create() {
|
2016-09-12 16:47:25 -07:00
|
|
|
webrtc::Config config;
|
2015-01-15 18:07:21 +00:00
|
|
|
return Create(config, nullptr);
|
2014-01-25 02:09:06 +00:00
|
|
|
}
|
|
|
|
|
|
2016-09-12 16:47:25 -07:00
|
|
|
AudioProcessing* AudioProcessing::Create(const webrtc::Config& config) {
|
2015-01-15 18:07:21 +00:00
|
|
|
return Create(config, nullptr);
|
|
|
|
|
}
|
|
|
|
|
|
2016-09-12 16:47:25 -07:00
|
|
|
AudioProcessing* AudioProcessing::Create(const webrtc::Config& config,
|
2016-07-01 17:19:09 -07:00
|
|
|
NonlinearBeamformer* beamformer) {
|
2015-01-15 18:07:21 +00:00
|
|
|
AudioProcessingImpl* apm = new AudioProcessingImpl(config, beamformer);
|
2011-07-07 08:21:25 +00:00
|
|
|
if (apm->Initialize() != kNoError) {
|
|
|
|
|
delete apm;
|
2015-11-28 12:35:15 -08:00
|
|
|
apm = nullptr;
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return apm;
|
|
|
|
|
}
|
|
|
|
|
|
2016-09-12 16:47:25 -07:00
|
|
|
AudioProcessingImpl::AudioProcessingImpl(const webrtc::Config& config)
|
2015-01-15 18:07:21 +00:00
|
|
|
: AudioProcessingImpl(config, nullptr) {}
|
|
|
|
|
|
2016-09-12 16:47:25 -07:00
|
|
|
AudioProcessingImpl::AudioProcessingImpl(const webrtc::Config& config,
|
2016-07-01 17:19:09 -07:00
|
|
|
NonlinearBeamformer* beamformer)
|
2016-11-22 07:24:52 -08:00
|
|
|
: high_pass_filter_impl_(new HighPassFilterImpl(this)),
|
|
|
|
|
public_submodules_(new ApmPublicSubmodules()),
|
2015-11-28 12:35:15 -08:00
|
|
|
private_submodules_(new ApmPrivateSubmodules(beamformer)),
|
|
|
|
|
constants_(config.Get<ExperimentalAgc>().startup_min_volume,
|
2014-12-15 09:41:24 +00:00
|
|
|
#if defined(WEBRTC_ANDROID) || defined(WEBRTC_IOS)
|
2016-05-16 15:32:38 -07:00
|
|
|
false),
|
2014-12-15 09:41:24 +00:00
|
|
|
#else
|
2016-05-16 15:32:38 -07:00
|
|
|
config.Get<ExperimentalAgc>().enabled),
|
2014-12-15 09:41:24 +00:00
|
|
|
#endif
|
2015-06-24 18:14:14 -07:00
|
|
|
#if defined(WEBRTC_ANDROID) || defined(WEBRTC_IOS)
|
2016-01-11 18:04:30 -08:00
|
|
|
capture_(false,
|
2015-06-24 18:14:14 -07:00
|
|
|
#else
|
2016-01-11 18:04:30 -08:00
|
|
|
capture_(config.Get<ExperimentalNs>().enabled,
|
2015-06-24 18:14:14 -07:00
|
|
|
#endif
|
2016-01-11 18:04:30 -08:00
|
|
|
config.Get<Beamforming>().array_geometry,
|
2016-01-11 20:32:29 -08:00
|
|
|
config.Get<Beamforming>().target_direction),
|
2016-05-16 15:32:38 -07:00
|
|
|
capture_nonlocked_(config.Get<Beamforming>().enabled,
|
2016-09-12 16:47:25 -07:00
|
|
|
config.Get<Intelligibility>().enabled) {
|
2015-11-28 12:35:15 -08:00
|
|
|
{
|
|
|
|
|
rtc::CritScope cs_render(&crit_render_);
|
|
|
|
|
rtc::CritScope cs_capture(&crit_capture_);
|
|
|
|
|
|
2016-03-05 03:01:14 -08:00
|
|
|
public_submodules_->echo_cancellation.reset(
|
2016-03-15 09:34:24 -07:00
|
|
|
new EchoCancellationImpl(&crit_render_, &crit_capture_));
|
2016-03-10 12:54:25 -08:00
|
|
|
public_submodules_->echo_control_mobile.reset(
|
2016-03-15 04:32:28 -07:00
|
|
|
new EchoControlMobileImpl(&crit_render_, &crit_capture_));
|
2016-03-10 21:09:04 -08:00
|
|
|
public_submodules_->gain_control.reset(
|
2016-03-15 02:28:08 -07:00
|
|
|
new GainControlImpl(&crit_capture_, &crit_capture_));
|
2015-12-15 11:39:38 -08:00
|
|
|
public_submodules_->level_estimator.reset(
|
|
|
|
|
new LevelEstimatorImpl(&crit_capture_));
|
2015-12-08 13:22:33 -08:00
|
|
|
public_submodules_->noise_suppression.reset(
|
|
|
|
|
new NoiseSuppressionImpl(&crit_capture_));
|
2015-12-16 03:31:12 -08:00
|
|
|
public_submodules_->voice_detection.reset(
|
|
|
|
|
new VoiceDetectionImpl(&crit_capture_));
|
2016-02-13 16:40:47 -08:00
|
|
|
public_submodules_->gain_control_for_experimental_agc.reset(
|
2016-03-10 21:09:04 -08:00
|
|
|
new GainControlForExperimentalAgc(
|
|
|
|
|
public_submodules_->gain_control.get(), &crit_capture_));
|
2016-10-28 05:39:16 -07:00
|
|
|
private_submodules_->residual_echo_detector.reset(
|
|
|
|
|
new ResidualEchoDetector());
|
2016-06-29 15:26:12 -07:00
|
|
|
|
2016-10-07 14:54:10 -07:00
|
|
|
// TODO(peah): Move this creation to happen only when the level controller
|
|
|
|
|
// is enabled.
|
2016-06-29 15:26:12 -07:00
|
|
|
private_submodules_->level_controller.reset(new LevelController());
|
2015-11-28 12:35:15 -08:00
|
|
|
}
|
2014-12-15 09:41:24 +00:00
|
|
|
|
2014-01-25 02:09:06 +00:00
|
|
|
SetExtraOptions(config);
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
AudioProcessingImpl::~AudioProcessingImpl() {
|
2015-11-28 12:35:15 -08:00
|
|
|
// Depends on gain_control_ and
|
2016-02-13 16:40:47 -08:00
|
|
|
// public_submodules_->gain_control_for_experimental_agc.
|
2015-11-28 12:35:15 -08:00
|
|
|
private_submodules_->agc_manager.reset();
|
|
|
|
|
// Depends on gain_control_.
|
2016-02-13 16:40:47 -08:00
|
|
|
public_submodules_->gain_control_for_experimental_agc.reset();
|
2011-07-07 08:21:25 +00:00
|
|
|
|
2011-12-03 00:03:31 +00:00
|
|
|
#ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP
|
FileWrapper[Impl] modifications and actually remove the "Impl" class.
This is a somewhat involved refactoring of this class. Here's an overview of the changes:
* FileWrapper can now be used as a regular class and instances allocated on the stack.
* The type now has support for move semantics and copy isn't allowed.
* New public ctor with FILE* that can be used instead of OpenFromFileHandle.
* New static Open() method. The intent of this is to allow opening a file and getting back a FileWrapper instance. Using this method instead of Create(), will allow us in the future to make the FILE* member pointer, to be const and simplify threading (get rid of the lock).
* Rename the Open() method to is_open() and make it inline.
* The FileWrapper interface is no longer a pure virtual interface. There's only one implementation so there's no need to go through a vtable for everything.
* Functionality offered by the class, is now reduced. No support for looping (not clear if that was actually useful to users of that flag), no need to implement the 'read_only_' functionality in the class, since file APIs implement that already, no support for *not* managing the file handle (this wasn't used). OpenFromFileHandle always "manages" the file.
* Delete the unused WriteText() method and don't support opening files in text mode. Text mode is only different on Windows and on Windows it translates \n to \r\n, which means that files such as log files, could have a slightly different format on Windows than other platforms. Besides, tools on Windows can handle UNIX line endings.
* Remove FileName(), change Trace code to manage its own path.
* Rename id_ member variable to file_.
* Removed the open_ member variable since the same functionality can be gotten from just checking the file pointer.
* Don't call CloseFile inside of Write. Write shouldn't be changing the state of the class beyond just attempting to write.
* Remove concept of looping from FileWrapper and never close inside of Read()
* Changed stream base classes to inherit from a common base class instead of both defining the Rewind method. Ultimately, Id' like to remove these interfaces and just have FileWrapper.
* Remove read_only param from OpenFromFileHandle
* Renamed size_in_bytes_ to position_, since it gets set to 0 when Rewind() is called (and the size actually does not change).
* Switch out rw lock for CriticalSection. The r/w lock was only used for reading when checking the open_ flag.
BUG=
Review-Url: https://codereview.webrtc.org/2054373002
Cr-Commit-Position: refs/heads/master@{#13155}
2016-06-15 10:30:14 -07:00
|
|
|
debug_dump_.debug_file->CloseFile();
|
2015-11-28 12:35:15 -08:00
|
|
|
#endif
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
int AudioProcessingImpl::Initialize() {
|
2015-11-28 12:35:15 -08:00
|
|
|
// Run in a single-threaded manner during initialization.
|
|
|
|
|
rtc::CritScope cs_render(&crit_render_);
|
|
|
|
|
rtc::CritScope cs_capture(&crit_capture_);
|
2011-07-07 08:21:25 +00:00
|
|
|
return InitializeLocked();
|
|
|
|
|
}
|
|
|
|
|
|
2016-09-16 15:02:15 -07:00
|
|
|
int AudioProcessingImpl::Initialize(int capture_input_sample_rate_hz,
|
|
|
|
|
int capture_output_sample_rate_hz,
|
|
|
|
|
int render_input_sample_rate_hz,
|
|
|
|
|
ChannelLayout capture_input_layout,
|
|
|
|
|
ChannelLayout capture_output_layout,
|
|
|
|
|
ChannelLayout render_input_layout) {
|
2015-07-23 11:41:39 -07:00
|
|
|
const ProcessingConfig processing_config = {
|
2016-09-16 15:02:15 -07:00
|
|
|
{{capture_input_sample_rate_hz, ChannelsFromLayout(capture_input_layout),
|
|
|
|
|
LayoutHasKeyboard(capture_input_layout)},
|
|
|
|
|
{capture_output_sample_rate_hz,
|
|
|
|
|
ChannelsFromLayout(capture_output_layout),
|
|
|
|
|
LayoutHasKeyboard(capture_output_layout)},
|
|
|
|
|
{render_input_sample_rate_hz, ChannelsFromLayout(render_input_layout),
|
|
|
|
|
LayoutHasKeyboard(render_input_layout)},
|
|
|
|
|
{render_input_sample_rate_hz, ChannelsFromLayout(render_input_layout),
|
|
|
|
|
LayoutHasKeyboard(render_input_layout)}}};
|
2015-07-23 11:41:39 -07:00
|
|
|
|
|
|
|
|
return Initialize(processing_config);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
int AudioProcessingImpl::Initialize(const ProcessingConfig& processing_config) {
|
2015-11-28 12:35:15 -08:00
|
|
|
// Run in a single-threaded manner during initialization.
|
|
|
|
|
rtc::CritScope cs_render(&crit_render_);
|
|
|
|
|
rtc::CritScope cs_capture(&crit_capture_);
|
2015-07-23 11:41:39 -07:00
|
|
|
return InitializeLocked(processing_config);
|
2014-03-10 22:26:12 +00:00
|
|
|
}
|
|
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
int AudioProcessingImpl::MaybeInitializeRender(
|
2015-11-27 02:47:28 -08:00
|
|
|
const ProcessingConfig& processing_config) {
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
return MaybeInitialize(processing_config, false);
|
2015-11-27 02:47:28 -08:00
|
|
|
}
|
|
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
int AudioProcessingImpl::MaybeInitializeCapture(
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
const ProcessingConfig& processing_config,
|
|
|
|
|
bool force_initialization) {
|
|
|
|
|
return MaybeInitialize(processing_config, force_initialization);
|
2015-11-27 02:47:28 -08:00
|
|
|
}
|
|
|
|
|
|
2016-08-29 14:46:07 -07:00
|
|
|
#ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP
|
|
|
|
|
|
|
|
|
|
AudioProcessingImpl::ApmDebugDumpThreadState::ApmDebugDumpThreadState()
|
|
|
|
|
: event_msg(new audioproc::Event()) {}
|
|
|
|
|
|
|
|
|
|
AudioProcessingImpl::ApmDebugDumpThreadState::~ApmDebugDumpThreadState() {}
|
|
|
|
|
|
|
|
|
|
AudioProcessingImpl::ApmDebugDumpState::ApmDebugDumpState()
|
|
|
|
|
: debug_file(FileWrapper::Create()) {}
|
|
|
|
|
|
|
|
|
|
AudioProcessingImpl::ApmDebugDumpState::~ApmDebugDumpState() {}
|
|
|
|
|
|
|
|
|
|
#endif // WEBRTC_AUDIOPROC_DEBUG_DUMP
|
|
|
|
|
|
2015-11-17 02:16:45 -08:00
|
|
|
// Calls InitializeLocked() if any of the audio parameters have changed from
|
2015-11-28 12:35:15 -08:00
|
|
|
// their current values (needs to be called while holding the crit_render_lock).
|
|
|
|
|
int AudioProcessingImpl::MaybeInitialize(
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
const ProcessingConfig& processing_config,
|
|
|
|
|
bool force_initialization) {
|
2015-11-28 12:35:15 -08:00
|
|
|
// Called from both threads. Thread check is therefore not possible.
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
if (processing_config == formats_.api_format && !force_initialization) {
|
2015-11-17 02:16:45 -08:00
|
|
|
return kNoError;
|
|
|
|
|
}
|
2015-11-28 12:35:15 -08:00
|
|
|
|
|
|
|
|
rtc::CritScope cs_capture(&crit_capture_);
|
2015-11-17 02:16:45 -08:00
|
|
|
return InitializeLocked(processing_config);
|
|
|
|
|
}
|
|
|
|
|
|
2011-07-07 08:21:25 +00:00
|
|
|
int AudioProcessingImpl::InitializeLocked() {
|
2016-09-16 15:02:15 -07:00
|
|
|
const int capture_audiobuffer_num_channels =
|
2016-01-11 20:32:29 -08:00
|
|
|
capture_nonlocked_.beamformer_enabled
|
2015-11-28 12:35:15 -08:00
|
|
|
? formats_.api_format.input_stream().num_channels()
|
|
|
|
|
: formats_.api_format.output_stream().num_channels();
|
2016-09-16 15:02:15 -07:00
|
|
|
const int render_audiobuffer_num_output_frames =
|
2015-11-28 12:35:15 -08:00
|
|
|
formats_.api_format.reverse_output_stream().num_frames() == 0
|
2016-09-16 15:02:15 -07:00
|
|
|
? formats_.render_processing_format.num_frames()
|
2015-11-28 12:35:15 -08:00
|
|
|
: formats_.api_format.reverse_output_stream().num_frames();
|
|
|
|
|
if (formats_.api_format.reverse_input_stream().num_channels() > 0) {
|
|
|
|
|
render_.render_audio.reset(new AudioBuffer(
|
|
|
|
|
formats_.api_format.reverse_input_stream().num_frames(),
|
|
|
|
|
formats_.api_format.reverse_input_stream().num_channels(),
|
2016-09-16 15:02:15 -07:00
|
|
|
formats_.render_processing_format.num_frames(),
|
|
|
|
|
formats_.render_processing_format.num_channels(),
|
|
|
|
|
render_audiobuffer_num_output_frames));
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
if (formats_.api_format.reverse_input_stream() !=
|
|
|
|
|
formats_.api_format.reverse_output_stream()) {
|
2016-02-24 05:22:32 -08:00
|
|
|
render_.render_converter = AudioConverter::Create(
|
2015-11-28 12:35:15 -08:00
|
|
|
formats_.api_format.reverse_input_stream().num_channels(),
|
|
|
|
|
formats_.api_format.reverse_input_stream().num_frames(),
|
|
|
|
|
formats_.api_format.reverse_output_stream().num_channels(),
|
2016-02-24 05:22:32 -08:00
|
|
|
formats_.api_format.reverse_output_stream().num_frames());
|
2015-08-14 10:35:55 -07:00
|
|
|
} else {
|
2015-11-28 12:35:15 -08:00
|
|
|
render_.render_converter.reset(nullptr);
|
2015-08-14 10:35:55 -07:00
|
|
|
}
|
2015-07-23 11:41:39 -07:00
|
|
|
} else {
|
2015-11-28 12:35:15 -08:00
|
|
|
render_.render_audio.reset(nullptr);
|
|
|
|
|
render_.render_converter.reset(nullptr);
|
2015-07-23 11:41:39 -07:00
|
|
|
}
|
2015-11-28 12:35:15 -08:00
|
|
|
capture_.capture_audio.reset(
|
|
|
|
|
new AudioBuffer(formats_.api_format.input_stream().num_frames(),
|
|
|
|
|
formats_.api_format.input_stream().num_channels(),
|
2016-09-16 15:02:15 -07:00
|
|
|
capture_nonlocked_.capture_processing_format.num_frames(),
|
|
|
|
|
capture_audiobuffer_num_channels,
|
2015-11-28 12:35:15 -08:00
|
|
|
formats_.api_format.output_stream().num_frames()));
|
2011-07-07 08:21:25 +00:00
|
|
|
|
2016-09-16 15:02:15 -07:00
|
|
|
public_submodules_->echo_cancellation->Initialize(
|
|
|
|
|
proc_sample_rate_hz(), num_reverse_channels(), num_output_channels(),
|
|
|
|
|
num_proc_channels());
|
2016-10-22 05:04:30 -07:00
|
|
|
AllocateRenderQueue();
|
|
|
|
|
|
2016-10-28 07:55:33 -07:00
|
|
|
int success = public_submodules_->echo_cancellation->enable_metrics(true);
|
|
|
|
|
RTC_DCHECK_EQ(0, success);
|
|
|
|
|
success = public_submodules_->echo_cancellation->enable_delay_logging(true);
|
|
|
|
|
RTC_DCHECK_EQ(0, success);
|
2016-09-16 15:02:15 -07:00
|
|
|
public_submodules_->echo_control_mobile->Initialize(
|
|
|
|
|
proc_split_sample_rate_hz(), num_reverse_channels(),
|
|
|
|
|
num_output_channels());
|
2016-10-28 03:12:11 -07:00
|
|
|
|
|
|
|
|
public_submodules_->gain_control->Initialize(num_proc_channels(),
|
|
|
|
|
proc_sample_rate_hz());
|
2016-09-16 15:02:15 -07:00
|
|
|
if (constants_.use_experimental_agc) {
|
|
|
|
|
if (!private_submodules_->agc_manager.get()) {
|
|
|
|
|
private_submodules_->agc_manager.reset(new AgcManagerDirect(
|
|
|
|
|
public_submodules_->gain_control.get(),
|
|
|
|
|
public_submodules_->gain_control_for_experimental_agc.get(),
|
|
|
|
|
constants_.agc_startup_min_volume));
|
|
|
|
|
}
|
|
|
|
|
private_submodules_->agc_manager->Initialize();
|
|
|
|
|
private_submodules_->agc_manager->SetCaptureMuted(
|
|
|
|
|
capture_.output_will_be_muted);
|
2016-10-28 03:12:11 -07:00
|
|
|
public_submodules_->gain_control_for_experimental_agc->Initialize();
|
2016-09-16 15:02:15 -07:00
|
|
|
}
|
2015-04-15 11:42:40 +02:00
|
|
|
InitializeTransient();
|
2014-12-19 19:57:34 +00:00
|
|
|
InitializeBeamformer();
|
2016-08-26 07:16:04 -07:00
|
|
|
#if WEBRTC_INTELLIGIBILITY_ENHANCER
|
2015-08-14 10:35:55 -07:00
|
|
|
InitializeIntelligibility();
|
2016-08-26 07:16:04 -07:00
|
|
|
#endif
|
2016-11-22 07:24:52 -08:00
|
|
|
InitializeLowCutFilter();
|
2016-09-16 15:02:15 -07:00
|
|
|
public_submodules_->noise_suppression->Initialize(num_proc_channels(),
|
|
|
|
|
proc_sample_rate_hz());
|
|
|
|
|
public_submodules_->voice_detection->Initialize(proc_split_sample_rate_hz());
|
|
|
|
|
public_submodules_->level_estimator->Initialize();
|
2016-06-29 15:26:12 -07:00
|
|
|
InitializeLevelController();
|
2016-10-28 05:39:16 -07:00
|
|
|
InitializeResidualEchoDetector();
|
2015-12-08 11:07:32 -08:00
|
|
|
|
2011-12-03 00:03:31 +00:00
|
|
|
#ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP
|
FileWrapper[Impl] modifications and actually remove the "Impl" class.
This is a somewhat involved refactoring of this class. Here's an overview of the changes:
* FileWrapper can now be used as a regular class and instances allocated on the stack.
* The type now has support for move semantics and copy isn't allowed.
* New public ctor with FILE* that can be used instead of OpenFromFileHandle.
* New static Open() method. The intent of this is to allow opening a file and getting back a FileWrapper instance. Using this method instead of Create(), will allow us in the future to make the FILE* member pointer, to be const and simplify threading (get rid of the lock).
* Rename the Open() method to is_open() and make it inline.
* The FileWrapper interface is no longer a pure virtual interface. There's only one implementation so there's no need to go through a vtable for everything.
* Functionality offered by the class, is now reduced. No support for looping (not clear if that was actually useful to users of that flag), no need to implement the 'read_only_' functionality in the class, since file APIs implement that already, no support for *not* managing the file handle (this wasn't used). OpenFromFileHandle always "manages" the file.
* Delete the unused WriteText() method and don't support opening files in text mode. Text mode is only different on Windows and on Windows it translates \n to \r\n, which means that files such as log files, could have a slightly different format on Windows than other platforms. Besides, tools on Windows can handle UNIX line endings.
* Remove FileName(), change Trace code to manage its own path.
* Rename id_ member variable to file_.
* Removed the open_ member variable since the same functionality can be gotten from just checking the file pointer.
* Don't call CloseFile inside of Write. Write shouldn't be changing the state of the class beyond just attempting to write.
* Remove concept of looping from FileWrapper and never close inside of Read()
* Changed stream base classes to inherit from a common base class instead of both defining the Rewind method. Ultimately, Id' like to remove these interfaces and just have FileWrapper.
* Remove read_only param from OpenFromFileHandle
* Renamed size_in_bytes_ to position_, since it gets set to 0 when Rewind() is called (and the size actually does not change).
* Switch out rw lock for CriticalSection. The r/w lock was only used for reading when checking the open_ flag.
BUG=
Review-Url: https://codereview.webrtc.org/2054373002
Cr-Commit-Position: refs/heads/master@{#13155}
2016-06-15 10:30:14 -07:00
|
|
|
if (debug_dump_.debug_file->is_open()) {
|
2011-08-03 21:08:51 +00:00
|
|
|
int err = WriteInitMessage();
|
|
|
|
|
if (err != kNoError) {
|
|
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
}
|
2011-12-03 00:03:31 +00:00
|
|
|
#endif
|
2011-08-03 21:08:51 +00:00
|
|
|
|
2011-07-07 08:21:25 +00:00
|
|
|
return kNoError;
|
|
|
|
|
}
|
|
|
|
|
|
2015-07-23 11:41:39 -07:00
|
|
|
int AudioProcessingImpl::InitializeLocked(const ProcessingConfig& config) {
|
|
|
|
|
for (const auto& stream : config.streams) {
|
|
|
|
|
if (stream.num_channels() > 0 && stream.sample_rate_hz() <= 0) {
|
|
|
|
|
return kBadSampleRateError;
|
|
|
|
|
}
|
Revert of Allow more than 2 input channels in AudioProcessing. (patchset #13 id:240001 of https://codereview.webrtc.org/1226093007/)
Reason for revert:
Breaks Chromium FYI content_browsertest on all platforms. The testcase that fails is WebRtcAecDumpBrowserTest.CallWithAecDump.
https://build.chromium.org/p/chromium.webrtc.fyi/builders/Linux/builds/19388
Sample output:
[ RUN ] WebRtcAecDumpBrowserTest.CallWithAecDump
Xlib: extension "RANDR" missing on display ":9".
[4:14:0722/211548:1282124453:WARNING:webrtcvoiceengine.cc(472)] Unexpected codec: ISAC/48000/1 (105)
[4:14:0722/211548:1282124593:WARNING:webrtcvoiceengine.cc(472)] Unexpected codec: PCMU/8000/2 (110)
[4:14:0722/211548:1282124700:WARNING:webrtcvoiceengine.cc(472)] Unexpected codec: PCMA/8000/2 (118)
[4:14:0722/211548:1282124815:WARNING:webrtcvoiceengine.cc(472)] Unexpected codec: G722/8000/2 (119)
[19745:19745:0722/211548:1282133667:INFO:CONSOLE(64)] "Looking at video in element remote-view-1", source: http://127.0.0.1:48819/media/webrtc_test_utilities.js (64)
[19745:19745:0722/211548:1282136892:INFO:CONSOLE(64)] "Looking at video in element remote-view-2", source: http://127.0.0.1:48819/media/webrtc_test_utilities.js (64)
../../content/test/webrtc_content_browsertest_base.cc:62: Failure
Value of: ExecuteScriptAndExtractString( shell()->web_contents(), javascript, &result)
Actual: false
Expected: true
Failed to execute javascript call({video: true, audio: true});.
From javascript: (nothing)
When executing 'call({video: true, audio: true});'
../../content/test/webrtc_content_browsertest_base.cc:75: Failure
Failed
../../content/browser/media/webrtc_aecdump_browsertest.cc:26: Failure
Expected: (base::kNullProcessId) != (*id), actual: 0 vs 0
../../content/browser/media/webrtc_aecdump_browsertest.cc:95: Failure
Value of: GetRenderProcessHostId(&render_process_id)
Actual: false
Expected: true
../../content/browser/media/webrtc_aecdump_browsertest.cc:99: Failure
Value of: base::PathExists(dump_file)
Actual: false
Expected: true
../../content/browser/media/webrtc_aecdump_browsertest.cc:101: Failure
Value of: base::GetFileSize(dump_file, &file_size)
Actual: false
Expected: true
../../content/browser/media/webrtc_aecdump_browsertest.cc:102: Failure
Expected: (file_size) > (0), actual: 0 vs 0
[ FAILED ] WebRtcAecDumpBrowserTest.CallWithAecDump, where TypeParam = and GetParam() = (361 ms)
Original issue's description:
> Allow more than 2 input channels in AudioProcessing.
>
> The number of output channels is constrained to be equal to either 1 or the
> number of input channels.
>
> R=aluebs@webrtc.org, andrew@webrtc.org, pbos@webrtc.org
>
> Committed: https://chromium.googlesource.com/external/webrtc/+/c204754b7a0cc801c70e8ce6c689f57f6ce00b3b
TBR=andrew@webrtc.org,aluebs@webrtc.org,ajm@chromium.org,pbos@chromium.org,pbos@webrtc.org,mgraczyk@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Review URL: https://codereview.webrtc.org/1253573005
Cr-Commit-Position: refs/heads/master@{#9621}
2015-07-23 04:30:06 -07:00
|
|
|
}
|
2015-07-23 11:41:39 -07:00
|
|
|
|
Convert channel counts to size_t.
IIRC, this was originally requested by ajm during review of the other size_t conversions I did over the past year, and I agreed it made sense, but wanted to do it separately since those changes were already gargantuan.
BUG=chromium:81439
TEST=none
R=henrik.lundin@webrtc.org, henrika@webrtc.org, kjellander@webrtc.org, minyue@webrtc.org, perkj@webrtc.org, solenberg@webrtc.org, stefan@webrtc.org, tina.legrand@webrtc.org
Review URL: https://codereview.webrtc.org/1316523002 .
Cr-Commit-Position: refs/heads/master@{#11229}
2016-01-12 16:26:35 -08:00
|
|
|
const size_t num_in_channels = config.input_stream().num_channels();
|
|
|
|
|
const size_t num_out_channels = config.output_stream().num_channels();
|
2015-07-23 11:41:39 -07:00
|
|
|
|
|
|
|
|
// Need at least one input channel.
|
|
|
|
|
// Need either one output channel or as many outputs as there are inputs.
|
|
|
|
|
if (num_in_channels == 0 ||
|
|
|
|
|
!(num_out_channels == 1 || num_out_channels == num_in_channels)) {
|
2014-03-10 22:26:12 +00:00
|
|
|
return kBadNumberChannelsError;
|
|
|
|
|
}
|
2015-07-23 11:41:39 -07:00
|
|
|
|
2016-01-11 20:32:29 -08:00
|
|
|
if (capture_nonlocked_.beamformer_enabled &&
|
Convert channel counts to size_t.
IIRC, this was originally requested by ajm during review of the other size_t conversions I did over the past year, and I agreed it made sense, but wanted to do it separately since those changes were already gargantuan.
BUG=chromium:81439
TEST=none
R=henrik.lundin@webrtc.org, henrika@webrtc.org, kjellander@webrtc.org, minyue@webrtc.org, perkj@webrtc.org, solenberg@webrtc.org, stefan@webrtc.org, tina.legrand@webrtc.org
Review URL: https://codereview.webrtc.org/1316523002 .
Cr-Commit-Position: refs/heads/master@{#11229}
2016-01-12 16:26:35 -08:00
|
|
|
num_in_channels != capture_.array_geometry.size()) {
|
2015-01-15 18:07:21 +00:00
|
|
|
return kBadNumberChannelsError;
|
|
|
|
|
}
|
2014-04-22 21:00:04 +00:00
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
formats_.api_format = config;
|
2014-04-22 21:00:04 +00:00
|
|
|
|
2016-09-16 15:02:15 -07:00
|
|
|
int capture_processing_rate = FindNativeProcessRateToUse(
|
2016-04-09 16:06:52 -07:00
|
|
|
std::min(formats_.api_format.input_stream().sample_rate_hz(),
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
formats_.api_format.output_stream().sample_rate_hz()),
|
|
|
|
|
submodule_states_.CaptureMultiBandSubModulesActive() ||
|
|
|
|
|
submodule_states_.RenderMultiBandSubModulesActive());
|
|
|
|
|
|
2016-09-16 15:02:15 -07:00
|
|
|
capture_nonlocked_.capture_processing_format =
|
|
|
|
|
StreamConfig(capture_processing_rate);
|
2014-04-22 21:00:04 +00:00
|
|
|
|
2016-09-16 15:02:15 -07:00
|
|
|
int render_processing_rate = FindNativeProcessRateToUse(
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
std::min(formats_.api_format.reverse_input_stream().sample_rate_hz(),
|
|
|
|
|
formats_.api_format.reverse_output_stream().sample_rate_hz()),
|
|
|
|
|
submodule_states_.CaptureMultiBandSubModulesActive() ||
|
|
|
|
|
submodule_states_.RenderMultiBandSubModulesActive());
|
2016-04-20 15:27:58 -07:00
|
|
|
// TODO(aluebs): Remove this restriction once we figure out why the 3-band
|
|
|
|
|
// splitting filter degrades the AEC performance.
|
2016-09-16 15:02:15 -07:00
|
|
|
if (render_processing_rate > kSampleRate32kHz) {
|
|
|
|
|
render_processing_rate = submodule_states_.RenderMultiBandProcessingActive()
|
|
|
|
|
? kSampleRate32kHz
|
|
|
|
|
: kSampleRate16kHz;
|
2016-04-20 15:27:58 -07:00
|
|
|
}
|
2016-09-16 15:02:15 -07:00
|
|
|
// If the forward sample rate is 8 kHz, the render stream is also processed
|
2016-04-20 15:27:58 -07:00
|
|
|
// at this rate.
|
2016-09-16 15:02:15 -07:00
|
|
|
if (capture_nonlocked_.capture_processing_format.sample_rate_hz() ==
|
|
|
|
|
kSampleRate8kHz) {
|
|
|
|
|
render_processing_rate = kSampleRate8kHz;
|
2014-04-22 21:00:04 +00:00
|
|
|
} else {
|
2016-09-16 15:02:15 -07:00
|
|
|
render_processing_rate =
|
|
|
|
|
std::max(render_processing_rate, static_cast<int>(kSampleRate16kHz));
|
2014-03-10 22:26:12 +00:00
|
|
|
}
|
|
|
|
|
|
2016-09-16 15:02:15 -07:00
|
|
|
// Always downmix the render stream to mono for analysis. This has been
|
Enable render downmixing to mono in AudioProcessing.
In practice, we have been doing this since time immemorial, but have
relied on the user to do the downmixing (first voice engine then
Chromium). It's more logical for this burden to fall on AudioProcessing,
however, who can be expected to know that this is a reasonable approach
for AEC. Permitting two render channels results in running two AECs
serially.
Critically, in my recent change to have Chromium adopt the float
interface:
https://codereview.chromium.org/420603004
I removed the downmixing by Chromium, forgetting that we hadn't yet
enabled this feature in AudioProcessing. This corrects that oversight.
The change in paths hit by production users is very minor. As commented
it required adding downmixing to the int16_t path to satisfy
bit-exactness tests.
For reference, find the ApmTest.Process errors here:
https://paste.googleplex.com/6372007910309888
BUG=webrtc:3853
TESTED=listened to the files output from the Process test, and verified
that they sound as expected: higher echo while the AEC is adapting, but
afterwards very close.
R=aluebs@webrtc.org, bjornv@webrtc.org, kwiberg@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/31459004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@7292 4adac7df-926f-26a2-2b94-8c16560cd09d
2014-09-24 20:06:23 +00:00
|
|
|
// demonstrated to work well for AEC in most practical scenarios.
|
2016-09-16 15:02:15 -07:00
|
|
|
formats_.render_processing_format = StreamConfig(render_processing_rate, 1);
|
2014-03-10 22:26:12 +00:00
|
|
|
|
2016-09-16 15:02:15 -07:00
|
|
|
if (capture_nonlocked_.capture_processing_format.sample_rate_hz() ==
|
|
|
|
|
kSampleRate32kHz ||
|
|
|
|
|
capture_nonlocked_.capture_processing_format.sample_rate_hz() ==
|
|
|
|
|
kSampleRate48kHz) {
|
2015-11-28 12:35:15 -08:00
|
|
|
capture_nonlocked_.split_rate = kSampleRate16kHz;
|
2014-03-10 22:26:12 +00:00
|
|
|
} else {
|
2015-11-28 12:35:15 -08:00
|
|
|
capture_nonlocked_.split_rate =
|
2016-09-16 15:02:15 -07:00
|
|
|
capture_nonlocked_.capture_processing_format.sample_rate_hz();
|
2014-03-10 22:26:12 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return InitializeLocked();
|
|
|
|
|
}
|
|
|
|
|
|
2016-09-12 16:47:25 -07:00
|
|
|
void AudioProcessingImpl::ApplyConfig(const AudioProcessing::Config& config) {
|
2016-10-07 14:54:10 -07:00
|
|
|
config_ = config;
|
2016-09-12 16:47:25 -07:00
|
|
|
|
2016-10-07 14:54:10 -07:00
|
|
|
bool config_ok = LevelController::Validate(config_.level_controller);
|
2016-09-12 16:47:25 -07:00
|
|
|
if (!config_ok) {
|
|
|
|
|
LOG(LS_ERROR) << "AudioProcessing module config error" << std::endl
|
|
|
|
|
<< "level_controller: "
|
2016-10-07 14:54:10 -07:00
|
|
|
<< LevelController::ToString(config_.level_controller)
|
2016-09-12 16:47:25 -07:00
|
|
|
<< std::endl
|
|
|
|
|
<< "Reverting to default parameter set";
|
2016-10-07 14:54:10 -07:00
|
|
|
config_.level_controller = AudioProcessing::Config::LevelController();
|
2016-09-12 16:47:25 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// Run in a single-threaded manner when applying the settings.
|
|
|
|
|
rtc::CritScope cs_render(&crit_render_);
|
|
|
|
|
rtc::CritScope cs_capture(&crit_capture_);
|
|
|
|
|
|
2016-10-07 14:54:10 -07:00
|
|
|
// TODO(peah): Replace the use of capture_nonlocked_.level_controller_enabled
|
|
|
|
|
// with the value in config_ everywhere in the code.
|
|
|
|
|
if (capture_nonlocked_.level_controller_enabled !=
|
|
|
|
|
config_.level_controller.enabled) {
|
2016-09-12 16:47:25 -07:00
|
|
|
capture_nonlocked_.level_controller_enabled =
|
2016-10-07 14:54:10 -07:00
|
|
|
config_.level_controller.enabled;
|
|
|
|
|
// TODO(peah): Remove the conditional initialization to always initialize
|
|
|
|
|
// the level controller regardless of whether it is enabled or not.
|
|
|
|
|
InitializeLevelController();
|
2016-09-12 16:47:25 -07:00
|
|
|
}
|
2016-10-07 14:54:10 -07:00
|
|
|
LOG(LS_INFO) << "Level controller activated: "
|
|
|
|
|
<< capture_nonlocked_.level_controller_enabled;
|
|
|
|
|
|
|
|
|
|
private_submodules_->level_controller->ApplyConfig(config_.level_controller);
|
2016-11-22 07:24:52 -08:00
|
|
|
|
|
|
|
|
InitializeLowCutFilter();
|
|
|
|
|
|
|
|
|
|
LOG(LS_INFO) << "Highpass filter activated: "
|
|
|
|
|
<< config_.high_pass_filter.enabled;
|
2016-09-12 16:47:25 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void AudioProcessingImpl::SetExtraOptions(const webrtc::Config& config) {
|
2015-11-28 12:35:15 -08:00
|
|
|
// Run in a single-threaded manner when setting the extra options.
|
|
|
|
|
rtc::CritScope cs_render(&crit_render_);
|
|
|
|
|
rtc::CritScope cs_capture(&crit_capture_);
|
2014-12-15 09:41:24 +00:00
|
|
|
|
2016-03-05 03:01:14 -08:00
|
|
|
public_submodules_->echo_cancellation->SetExtraOptions(config);
|
|
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
if (capture_.transient_suppressor_enabled !=
|
|
|
|
|
config.Get<ExperimentalNs>().enabled) {
|
|
|
|
|
capture_.transient_suppressor_enabled =
|
|
|
|
|
config.Get<ExperimentalNs>().enabled;
|
2014-12-15 09:41:24 +00:00
|
|
|
InitializeTransient();
|
|
|
|
|
}
|
2016-01-11 18:04:30 -08:00
|
|
|
|
2016-08-26 07:16:04 -07:00
|
|
|
#if WEBRTC_INTELLIGIBILITY_ENHANCER
|
2016-05-16 15:32:38 -07:00
|
|
|
if(capture_nonlocked_.intelligibility_enabled !=
|
|
|
|
|
config.Get<Intelligibility>().enabled) {
|
|
|
|
|
capture_nonlocked_.intelligibility_enabled =
|
|
|
|
|
config.Get<Intelligibility>().enabled;
|
|
|
|
|
InitializeIntelligibility();
|
|
|
|
|
}
|
2016-08-26 07:16:04 -07:00
|
|
|
#endif
|
2016-05-16 15:32:38 -07:00
|
|
|
|
2016-01-11 18:04:30 -08:00
|
|
|
#ifdef WEBRTC_ANDROID_PLATFORM_BUILD
|
2016-01-11 20:32:29 -08:00
|
|
|
if (capture_nonlocked_.beamformer_enabled !=
|
|
|
|
|
config.Get<Beamforming>().enabled) {
|
|
|
|
|
capture_nonlocked_.beamformer_enabled = config.Get<Beamforming>().enabled;
|
2016-01-11 18:04:30 -08:00
|
|
|
if (config.Get<Beamforming>().array_geometry.size() > 1) {
|
|
|
|
|
capture_.array_geometry = config.Get<Beamforming>().array_geometry;
|
|
|
|
|
}
|
|
|
|
|
capture_.target_direction = config.Get<Beamforming>().target_direction;
|
|
|
|
|
InitializeBeamformer();
|
|
|
|
|
}
|
|
|
|
|
#endif // WEBRTC_ANDROID_PLATFORM_BUILD
|
2013-07-25 18:28:29 +00:00
|
|
|
}
|
|
|
|
|
|
2014-04-22 21:00:04 +00:00
|
|
|
int AudioProcessingImpl::proc_sample_rate_hz() const {
|
2015-11-28 12:35:15 -08:00
|
|
|
// Used as callback from submodules, hence locking is not allowed.
|
2016-09-16 15:02:15 -07:00
|
|
|
return capture_nonlocked_.capture_processing_format.sample_rate_hz();
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
2014-04-22 21:00:04 +00:00
|
|
|
int AudioProcessingImpl::proc_split_sample_rate_hz() const {
|
2015-11-28 12:35:15 -08:00
|
|
|
// Used as callback from submodules, hence locking is not allowed.
|
|
|
|
|
return capture_nonlocked_.split_rate;
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
Convert channel counts to size_t.
IIRC, this was originally requested by ajm during review of the other size_t conversions I did over the past year, and I agreed it made sense, but wanted to do it separately since those changes were already gargantuan.
BUG=chromium:81439
TEST=none
R=henrik.lundin@webrtc.org, henrika@webrtc.org, kjellander@webrtc.org, minyue@webrtc.org, perkj@webrtc.org, solenberg@webrtc.org, stefan@webrtc.org, tina.legrand@webrtc.org
Review URL: https://codereview.webrtc.org/1316523002 .
Cr-Commit-Position: refs/heads/master@{#11229}
2016-01-12 16:26:35 -08:00
|
|
|
size_t AudioProcessingImpl::num_reverse_channels() const {
|
2015-11-28 12:35:15 -08:00
|
|
|
// Used as callback from submodules, hence locking is not allowed.
|
2016-09-16 15:02:15 -07:00
|
|
|
return formats_.render_processing_format.num_channels();
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
Convert channel counts to size_t.
IIRC, this was originally requested by ajm during review of the other size_t conversions I did over the past year, and I agreed it made sense, but wanted to do it separately since those changes were already gargantuan.
BUG=chromium:81439
TEST=none
R=henrik.lundin@webrtc.org, henrika@webrtc.org, kjellander@webrtc.org, minyue@webrtc.org, perkj@webrtc.org, solenberg@webrtc.org, stefan@webrtc.org, tina.legrand@webrtc.org
Review URL: https://codereview.webrtc.org/1316523002 .
Cr-Commit-Position: refs/heads/master@{#11229}
2016-01-12 16:26:35 -08:00
|
|
|
size_t AudioProcessingImpl::num_input_channels() const {
|
2015-11-28 12:35:15 -08:00
|
|
|
// Used as callback from submodules, hence locking is not allowed.
|
|
|
|
|
return formats_.api_format.input_stream().num_channels();
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
Convert channel counts to size_t.
IIRC, this was originally requested by ajm during review of the other size_t conversions I did over the past year, and I agreed it made sense, but wanted to do it separately since those changes were already gargantuan.
BUG=chromium:81439
TEST=none
R=henrik.lundin@webrtc.org, henrika@webrtc.org, kjellander@webrtc.org, minyue@webrtc.org, perkj@webrtc.org, solenberg@webrtc.org, stefan@webrtc.org, tina.legrand@webrtc.org
Review URL: https://codereview.webrtc.org/1316523002 .
Cr-Commit-Position: refs/heads/master@{#11229}
2016-01-12 16:26:35 -08:00
|
|
|
size_t AudioProcessingImpl::num_proc_channels() const {
|
2016-01-11 20:32:29 -08:00
|
|
|
// Used as callback from submodules, hence locking is not allowed.
|
|
|
|
|
return capture_nonlocked_.beamformer_enabled ? 1 : num_output_channels();
|
|
|
|
|
}
|
|
|
|
|
|
Convert channel counts to size_t.
IIRC, this was originally requested by ajm during review of the other size_t conversions I did over the past year, and I agreed it made sense, but wanted to do it separately since those changes were already gargantuan.
BUG=chromium:81439
TEST=none
R=henrik.lundin@webrtc.org, henrika@webrtc.org, kjellander@webrtc.org, minyue@webrtc.org, perkj@webrtc.org, solenberg@webrtc.org, stefan@webrtc.org, tina.legrand@webrtc.org
Review URL: https://codereview.webrtc.org/1316523002 .
Cr-Commit-Position: refs/heads/master@{#11229}
2016-01-12 16:26:35 -08:00
|
|
|
size_t AudioProcessingImpl::num_output_channels() const {
|
2015-11-28 12:35:15 -08:00
|
|
|
// Used as callback from submodules, hence locking is not allowed.
|
|
|
|
|
return formats_.api_format.output_stream().num_channels();
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
2014-02-12 22:28:31 +00:00
|
|
|
void AudioProcessingImpl::set_output_will_be_muted(bool muted) {
|
2015-11-28 12:35:15 -08:00
|
|
|
rtc::CritScope cs(&crit_capture_);
|
|
|
|
|
capture_.output_will_be_muted = muted;
|
|
|
|
|
if (private_submodules_->agc_manager.get()) {
|
|
|
|
|
private_submodules_->agc_manager->SetCaptureMuted(
|
|
|
|
|
capture_.output_will_be_muted);
|
2014-12-15 09:41:24 +00:00
|
|
|
}
|
2014-02-12 22:28:31 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2014-04-22 21:00:04 +00:00
|
|
|
int AudioProcessingImpl::ProcessStream(const float* const* src,
|
Update a ton of audio code to use size_t more correctly and in general reduce
use of int16_t/uint16_t.
This is the upshot of a recommendation by henrik.lundin and kwiberg on an original small change ( https://webrtc-codereview.appspot.com/42569004/#ps1 ) to stop using int16_t just because values could fit in it, and is similar in nature to a previous "mass change to use size_t more" ( https://webrtc-codereview.appspot.com/23129004/ ) which also needed to be split up for review but to land all at once, since, like adding "const", such changes tend to cause a lot of transitive effects.
This was be reviewed and approved in pieces:
https://codereview.webrtc.org/1224093003
https://codereview.webrtc.org/1224123002
https://codereview.webrtc.org/1224163002
https://codereview.webrtc.org/1225133003
https://codereview.webrtc.org/1225173002
https://codereview.webrtc.org/1227163003
https://codereview.webrtc.org/1227203003
https://codereview.webrtc.org/1227213002
https://codereview.webrtc.org/1227893002
https://codereview.webrtc.org/1228793004
https://codereview.webrtc.org/1228803003
https://codereview.webrtc.org/1228823002
https://codereview.webrtc.org/1228823003
https://codereview.webrtc.org/1228843002
https://codereview.webrtc.org/1230693002
https://codereview.webrtc.org/1231713002
The change is being landed as TBR to all the folks who reviewed the above.
BUG=chromium:81439
TEST=none
R=andrew@webrtc.org, pbos@webrtc.org
TBR=aluebs, andrew, asapersson, henrika, hlundin, jan.skoglund, kwiberg, minyue, pbos, pthatcher
Review URL: https://codereview.webrtc.org/1230503003 .
Cr-Commit-Position: refs/heads/master@{#9768}
2015-08-24 14:52:23 -07:00
|
|
|
size_t samples_per_channel,
|
2014-04-22 21:00:04 +00:00
|
|
|
int input_sample_rate_hz,
|
2014-03-04 20:58:13 +00:00
|
|
|
ChannelLayout input_layout,
|
2014-04-22 21:00:04 +00:00
|
|
|
int output_sample_rate_hz,
|
|
|
|
|
ChannelLayout output_layout,
|
|
|
|
|
float* const* dest) {
|
2015-12-17 06:42:29 -08:00
|
|
|
TRACE_EVENT0("webrtc", "AudioProcessing::ProcessStream_ChannelLayout");
|
2015-11-28 12:35:15 -08:00
|
|
|
StreamConfig input_stream;
|
|
|
|
|
StreamConfig output_stream;
|
|
|
|
|
{
|
|
|
|
|
// Access the formats_.api_format.input_stream beneath the capture lock.
|
|
|
|
|
// The lock must be released as it is later required in the call
|
|
|
|
|
// to ProcessStream(,,,);
|
|
|
|
|
rtc::CritScope cs(&crit_capture_);
|
|
|
|
|
input_stream = formats_.api_format.input_stream();
|
|
|
|
|
output_stream = formats_.api_format.output_stream();
|
|
|
|
|
}
|
|
|
|
|
|
2015-07-23 11:41:39 -07:00
|
|
|
input_stream.set_sample_rate_hz(input_sample_rate_hz);
|
|
|
|
|
input_stream.set_num_channels(ChannelsFromLayout(input_layout));
|
|
|
|
|
input_stream.set_has_keyboard(LayoutHasKeyboard(input_layout));
|
|
|
|
|
output_stream.set_sample_rate_hz(output_sample_rate_hz);
|
|
|
|
|
output_stream.set_num_channels(ChannelsFromLayout(output_layout));
|
|
|
|
|
output_stream.set_has_keyboard(LayoutHasKeyboard(output_layout));
|
|
|
|
|
|
|
|
|
|
if (samples_per_channel != input_stream.num_frames()) {
|
|
|
|
|
return kBadDataLengthError;
|
|
|
|
|
}
|
|
|
|
|
return ProcessStream(src, input_stream, output_stream, dest);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
int AudioProcessingImpl::ProcessStream(const float* const* src,
|
|
|
|
|
const StreamConfig& input_config,
|
|
|
|
|
const StreamConfig& output_config,
|
|
|
|
|
float* const* dest) {
|
2015-12-17 06:42:29 -08:00
|
|
|
TRACE_EVENT0("webrtc", "AudioProcessing::ProcessStream_StreamConfig");
|
2015-11-28 12:35:15 -08:00
|
|
|
ProcessingConfig processing_config;
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
bool reinitialization_required = false;
|
2015-11-28 12:35:15 -08:00
|
|
|
{
|
|
|
|
|
// Acquire the capture lock in order to safely call the function
|
|
|
|
|
// that retrieves the render side data. This function accesses apm
|
|
|
|
|
// getters that need the capture lock held when being called.
|
|
|
|
|
rtc::CritScope cs_capture(&crit_capture_);
|
2016-10-22 05:04:30 -07:00
|
|
|
EmptyQueuedRenderAudio();
|
2015-11-28 12:35:15 -08:00
|
|
|
|
|
|
|
|
if (!src || !dest) {
|
|
|
|
|
return kNullPointerError;
|
|
|
|
|
}
|
2014-01-07 17:45:09 +00:00
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
processing_config = formats_.api_format;
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
reinitialization_required = UpdateActiveSubmoduleStates();
|
2015-11-28 12:35:15 -08:00
|
|
|
}
|
2015-11-16 16:27:42 -08:00
|
|
|
|
2015-07-23 11:41:39 -07:00
|
|
|
processing_config.input_stream() = input_config;
|
|
|
|
|
processing_config.output_stream() = output_config;
|
|
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
{
|
|
|
|
|
// Do conditional reinitialization.
|
|
|
|
|
rtc::CritScope cs_render(&crit_render_);
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
RETURN_ON_ERR(
|
|
|
|
|
MaybeInitializeCapture(processing_config, reinitialization_required));
|
2015-11-28 12:35:15 -08:00
|
|
|
}
|
|
|
|
|
rtc::CritScope cs_capture(&crit_capture_);
|
2016-09-14 05:23:22 -07:00
|
|
|
RTC_DCHECK_EQ(processing_config.input_stream().num_frames(),
|
|
|
|
|
formats_.api_format.input_stream().num_frames());
|
2014-03-04 20:58:13 +00:00
|
|
|
|
|
|
|
|
#ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP
|
FileWrapper[Impl] modifications and actually remove the "Impl" class.
This is a somewhat involved refactoring of this class. Here's an overview of the changes:
* FileWrapper can now be used as a regular class and instances allocated on the stack.
* The type now has support for move semantics and copy isn't allowed.
* New public ctor with FILE* that can be used instead of OpenFromFileHandle.
* New static Open() method. The intent of this is to allow opening a file and getting back a FileWrapper instance. Using this method instead of Create(), will allow us in the future to make the FILE* member pointer, to be const and simplify threading (get rid of the lock).
* Rename the Open() method to is_open() and make it inline.
* The FileWrapper interface is no longer a pure virtual interface. There's only one implementation so there's no need to go through a vtable for everything.
* Functionality offered by the class, is now reduced. No support for looping (not clear if that was actually useful to users of that flag), no need to implement the 'read_only_' functionality in the class, since file APIs implement that already, no support for *not* managing the file handle (this wasn't used). OpenFromFileHandle always "manages" the file.
* Delete the unused WriteText() method and don't support opening files in text mode. Text mode is only different on Windows and on Windows it translates \n to \r\n, which means that files such as log files, could have a slightly different format on Windows than other platforms. Besides, tools on Windows can handle UNIX line endings.
* Remove FileName(), change Trace code to manage its own path.
* Rename id_ member variable to file_.
* Removed the open_ member variable since the same functionality can be gotten from just checking the file pointer.
* Don't call CloseFile inside of Write. Write shouldn't be changing the state of the class beyond just attempting to write.
* Remove concept of looping from FileWrapper and never close inside of Read()
* Changed stream base classes to inherit from a common base class instead of both defining the Rewind method. Ultimately, Id' like to remove these interfaces and just have FileWrapper.
* Remove read_only param from OpenFromFileHandle
* Renamed size_in_bytes_ to position_, since it gets set to 0 when Rewind() is called (and the size actually does not change).
* Switch out rw lock for CriticalSection. The r/w lock was only used for reading when checking the open_ flag.
BUG=
Review-Url: https://codereview.webrtc.org/2054373002
Cr-Commit-Position: refs/heads/master@{#13155}
2016-06-15 10:30:14 -07:00
|
|
|
if (debug_dump_.debug_file->is_open()) {
|
2015-10-03 00:39:14 +02:00
|
|
|
RETURN_ON_ERR(WriteConfigMessage(false));
|
|
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
debug_dump_.capture.event_msg->set_type(audioproc::Event::STREAM);
|
|
|
|
|
audioproc::Stream* msg = debug_dump_.capture.event_msg->mutable_stream();
|
2014-08-28 10:43:09 +00:00
|
|
|
const size_t channel_size =
|
2015-11-28 12:35:15 -08:00
|
|
|
sizeof(float) * formats_.api_format.input_stream().num_frames();
|
Convert channel counts to size_t.
IIRC, this was originally requested by ajm during review of the other size_t conversions I did over the past year, and I agreed it made sense, but wanted to do it separately since those changes were already gargantuan.
BUG=chromium:81439
TEST=none
R=henrik.lundin@webrtc.org, henrika@webrtc.org, kjellander@webrtc.org, minyue@webrtc.org, perkj@webrtc.org, solenberg@webrtc.org, stefan@webrtc.org, tina.legrand@webrtc.org
Review URL: https://codereview.webrtc.org/1316523002 .
Cr-Commit-Position: refs/heads/master@{#11229}
2016-01-12 16:26:35 -08:00
|
|
|
for (size_t i = 0; i < formats_.api_format.input_stream().num_channels();
|
|
|
|
|
++i)
|
2014-04-22 21:00:04 +00:00
|
|
|
msg->add_input_channel(src[i], channel_size);
|
2014-03-04 20:58:13 +00:00
|
|
|
}
|
|
|
|
|
#endif
|
|
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
capture_.capture_audio->CopyFrom(src, formats_.api_format.input_stream());
|
2016-09-16 15:02:15 -07:00
|
|
|
RETURN_ON_ERR(ProcessCaptureStreamLocked());
|
2015-11-28 12:35:15 -08:00
|
|
|
capture_.capture_audio->CopyTo(formats_.api_format.output_stream(), dest);
|
2014-03-04 20:58:13 +00:00
|
|
|
|
|
|
|
|
#ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP
|
FileWrapper[Impl] modifications and actually remove the "Impl" class.
This is a somewhat involved refactoring of this class. Here's an overview of the changes:
* FileWrapper can now be used as a regular class and instances allocated on the stack.
* The type now has support for move semantics and copy isn't allowed.
* New public ctor with FILE* that can be used instead of OpenFromFileHandle.
* New static Open() method. The intent of this is to allow opening a file and getting back a FileWrapper instance. Using this method instead of Create(), will allow us in the future to make the FILE* member pointer, to be const and simplify threading (get rid of the lock).
* Rename the Open() method to is_open() and make it inline.
* The FileWrapper interface is no longer a pure virtual interface. There's only one implementation so there's no need to go through a vtable for everything.
* Functionality offered by the class, is now reduced. No support for looping (not clear if that was actually useful to users of that flag), no need to implement the 'read_only_' functionality in the class, since file APIs implement that already, no support for *not* managing the file handle (this wasn't used). OpenFromFileHandle always "manages" the file.
* Delete the unused WriteText() method and don't support opening files in text mode. Text mode is only different on Windows and on Windows it translates \n to \r\n, which means that files such as log files, could have a slightly different format on Windows than other platforms. Besides, tools on Windows can handle UNIX line endings.
* Remove FileName(), change Trace code to manage its own path.
* Rename id_ member variable to file_.
* Removed the open_ member variable since the same functionality can be gotten from just checking the file pointer.
* Don't call CloseFile inside of Write. Write shouldn't be changing the state of the class beyond just attempting to write.
* Remove concept of looping from FileWrapper and never close inside of Read()
* Changed stream base classes to inherit from a common base class instead of both defining the Rewind method. Ultimately, Id' like to remove these interfaces and just have FileWrapper.
* Remove read_only param from OpenFromFileHandle
* Renamed size_in_bytes_ to position_, since it gets set to 0 when Rewind() is called (and the size actually does not change).
* Switch out rw lock for CriticalSection. The r/w lock was only used for reading when checking the open_ flag.
BUG=
Review-Url: https://codereview.webrtc.org/2054373002
Cr-Commit-Position: refs/heads/master@{#13155}
2016-06-15 10:30:14 -07:00
|
|
|
if (debug_dump_.debug_file->is_open()) {
|
2015-11-28 12:35:15 -08:00
|
|
|
audioproc::Stream* msg = debug_dump_.capture.event_msg->mutable_stream();
|
2014-08-28 10:43:09 +00:00
|
|
|
const size_t channel_size =
|
2015-11-28 12:35:15 -08:00
|
|
|
sizeof(float) * formats_.api_format.output_stream().num_frames();
|
Convert channel counts to size_t.
IIRC, this was originally requested by ajm during review of the other size_t conversions I did over the past year, and I agreed it made sense, but wanted to do it separately since those changes were already gargantuan.
BUG=chromium:81439
TEST=none
R=henrik.lundin@webrtc.org, henrika@webrtc.org, kjellander@webrtc.org, minyue@webrtc.org, perkj@webrtc.org, solenberg@webrtc.org, stefan@webrtc.org, tina.legrand@webrtc.org
Review URL: https://codereview.webrtc.org/1316523002 .
Cr-Commit-Position: refs/heads/master@{#11229}
2016-01-12 16:26:35 -08:00
|
|
|
for (size_t i = 0; i < formats_.api_format.output_stream().num_channels();
|
|
|
|
|
++i)
|
2014-04-22 21:00:04 +00:00
|
|
|
msg->add_output_channel(dest[i], channel_size);
|
2015-11-28 12:35:15 -08:00
|
|
|
RETURN_ON_ERR(WriteMessageToDebugFile(debug_dump_.debug_file.get(),
|
2016-01-15 03:06:36 -08:00
|
|
|
&debug_dump_.num_bytes_left_for_log_,
|
2015-11-28 12:35:15 -08:00
|
|
|
&crit_debug_, &debug_dump_.capture));
|
2014-03-04 20:58:13 +00:00
|
|
|
}
|
|
|
|
|
#endif
|
|
|
|
|
|
|
|
|
|
return kNoError;
|
|
|
|
|
}
|
|
|
|
|
|
2016-10-25 05:42:20 -07:00
|
|
|
void AudioProcessingImpl::QueueRenderAudio(AudioBuffer* audio) {
|
2016-10-22 05:04:30 -07:00
|
|
|
EchoCancellationImpl::PackRenderAudioBuffer(audio, num_output_channels(),
|
|
|
|
|
num_reverse_channels(),
|
2016-10-25 05:42:20 -07:00
|
|
|
&aec_render_queue_buffer_);
|
2016-10-22 05:04:30 -07:00
|
|
|
|
2016-11-28 15:21:39 -08:00
|
|
|
RTC_DCHECK_GE(160, audio->num_frames_per_band());
|
2016-10-22 05:04:30 -07:00
|
|
|
|
|
|
|
|
// Insert the samples into the queue.
|
2016-10-25 05:42:20 -07:00
|
|
|
if (!aec_render_signal_queue_->Insert(&aec_render_queue_buffer_)) {
|
2016-10-22 05:04:30 -07:00
|
|
|
// The data queue is full and needs to be emptied.
|
|
|
|
|
EmptyQueuedRenderAudio();
|
|
|
|
|
|
|
|
|
|
// Retry the insert (should always work).
|
2016-10-25 05:42:20 -07:00
|
|
|
bool result = aec_render_signal_queue_->Insert(&aec_render_queue_buffer_);
|
2016-10-25 04:45:24 -07:00
|
|
|
RTC_DCHECK(result);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
EchoControlMobileImpl::PackRenderAudioBuffer(audio, num_output_channels(),
|
|
|
|
|
num_reverse_channels(),
|
2016-10-25 05:42:20 -07:00
|
|
|
&aecm_render_queue_buffer_);
|
2016-10-25 04:45:24 -07:00
|
|
|
|
|
|
|
|
// Insert the samples into the queue.
|
2016-10-25 05:42:20 -07:00
|
|
|
if (!aecm_render_signal_queue_->Insert(&aecm_render_queue_buffer_)) {
|
2016-10-25 04:45:24 -07:00
|
|
|
// The data queue is full and needs to be emptied.
|
|
|
|
|
EmptyQueuedRenderAudio();
|
|
|
|
|
|
|
|
|
|
// Retry the insert (should always work).
|
2016-10-25 05:42:20 -07:00
|
|
|
bool result = aecm_render_signal_queue_->Insert(&aecm_render_queue_buffer_);
|
2016-10-22 05:04:30 -07:00
|
|
|
RTC_DCHECK(result);
|
|
|
|
|
}
|
2016-10-25 05:42:20 -07:00
|
|
|
|
|
|
|
|
if (!constants_.use_experimental_agc) {
|
|
|
|
|
GainControlImpl::PackRenderAudioBuffer(audio, &agc_render_queue_buffer_);
|
|
|
|
|
// Insert the samples into the queue.
|
|
|
|
|
if (!agc_render_signal_queue_->Insert(&agc_render_queue_buffer_)) {
|
|
|
|
|
// The data queue is full and needs to be emptied.
|
|
|
|
|
EmptyQueuedRenderAudio();
|
|
|
|
|
|
|
|
|
|
// Retry the insert (should always work).
|
|
|
|
|
bool result = agc_render_signal_queue_->Insert(&agc_render_queue_buffer_);
|
|
|
|
|
RTC_DCHECK(result);
|
|
|
|
|
}
|
|
|
|
|
}
|
2016-10-28 05:39:16 -07:00
|
|
|
|
|
|
|
|
ResidualEchoDetector::PackRenderAudioBuffer(audio, &red_render_queue_buffer_);
|
|
|
|
|
|
|
|
|
|
// Insert the samples into the queue.
|
|
|
|
|
if (!red_render_signal_queue_->Insert(&red_render_queue_buffer_)) {
|
|
|
|
|
// The data queue is full and needs to be emptied.
|
|
|
|
|
EmptyQueuedRenderAudio();
|
|
|
|
|
|
|
|
|
|
// Retry the insert (should always work).
|
|
|
|
|
bool result = red_render_signal_queue_->Insert(&red_render_queue_buffer_);
|
|
|
|
|
RTC_DCHECK(result);
|
|
|
|
|
}
|
2016-10-22 05:04:30 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void AudioProcessingImpl::AllocateRenderQueue() {
|
2016-10-25 05:42:20 -07:00
|
|
|
const size_t new_aec_render_queue_element_max_size =
|
2016-10-22 05:04:30 -07:00
|
|
|
std::max(static_cast<size_t>(1),
|
|
|
|
|
kMaxAllowedValuesOfSamplesPerFrame *
|
|
|
|
|
EchoCancellationImpl::NumCancellersRequired(
|
|
|
|
|
num_output_channels(), num_reverse_channels()));
|
|
|
|
|
|
2016-10-25 05:42:20 -07:00
|
|
|
const size_t new_aecm_render_queue_element_max_size =
|
2016-10-25 04:45:24 -07:00
|
|
|
std::max(static_cast<size_t>(1),
|
|
|
|
|
kMaxAllowedValuesOfSamplesPerFrame *
|
|
|
|
|
EchoControlMobileImpl::NumCancellersRequired(
|
|
|
|
|
num_output_channels(), num_reverse_channels()));
|
|
|
|
|
|
2016-10-25 05:42:20 -07:00
|
|
|
const size_t new_agc_render_queue_element_max_size =
|
|
|
|
|
std::max(static_cast<size_t>(1), kMaxAllowedValuesOfSamplesPerFrame);
|
|
|
|
|
|
2016-10-28 05:39:16 -07:00
|
|
|
const size_t new_red_render_queue_element_max_size =
|
|
|
|
|
std::max(static_cast<size_t>(1), kMaxAllowedValuesOfSamplesPerFrame);
|
|
|
|
|
|
2016-10-25 04:45:24 -07:00
|
|
|
// Reallocate the queues if the queue item sizes are too small to fit the
|
|
|
|
|
// data to put in the queues.
|
2016-10-25 05:42:20 -07:00
|
|
|
if (aec_render_queue_element_max_size_ <
|
|
|
|
|
new_aec_render_queue_element_max_size) {
|
|
|
|
|
aec_render_queue_element_max_size_ = new_aec_render_queue_element_max_size;
|
2016-10-22 05:04:30 -07:00
|
|
|
|
2016-10-25 04:45:24 -07:00
|
|
|
std::vector<float> template_queue_element(
|
2016-10-25 05:42:20 -07:00
|
|
|
aec_render_queue_element_max_size_);
|
2016-10-22 05:04:30 -07:00
|
|
|
|
2016-10-25 05:42:20 -07:00
|
|
|
aec_render_signal_queue_.reset(
|
2016-10-22 05:04:30 -07:00
|
|
|
new SwapQueue<std::vector<float>, RenderQueueItemVerifier<float>>(
|
|
|
|
|
kMaxNumFramesToBuffer, template_queue_element,
|
2016-10-25 04:45:24 -07:00
|
|
|
RenderQueueItemVerifier<float>(
|
2016-10-25 05:42:20 -07:00
|
|
|
aec_render_queue_element_max_size_)));
|
|
|
|
|
|
|
|
|
|
aec_render_queue_buffer_.resize(aec_render_queue_element_max_size_);
|
|
|
|
|
aec_capture_queue_buffer_.resize(aec_render_queue_element_max_size_);
|
|
|
|
|
} else {
|
|
|
|
|
aec_render_signal_queue_->Clear();
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (aecm_render_queue_element_max_size_ <
|
|
|
|
|
new_aecm_render_queue_element_max_size) {
|
|
|
|
|
aecm_render_queue_element_max_size_ =
|
|
|
|
|
new_aecm_render_queue_element_max_size;
|
|
|
|
|
|
|
|
|
|
std::vector<int16_t> template_queue_element(
|
|
|
|
|
aecm_render_queue_element_max_size_);
|
|
|
|
|
|
|
|
|
|
aecm_render_signal_queue_.reset(
|
|
|
|
|
new SwapQueue<std::vector<int16_t>, RenderQueueItemVerifier<int16_t>>(
|
|
|
|
|
kMaxNumFramesToBuffer, template_queue_element,
|
|
|
|
|
RenderQueueItemVerifier<int16_t>(
|
|
|
|
|
aecm_render_queue_element_max_size_)));
|
2016-10-22 05:04:30 -07:00
|
|
|
|
2016-10-25 05:42:20 -07:00
|
|
|
aecm_render_queue_buffer_.resize(aecm_render_queue_element_max_size_);
|
|
|
|
|
aecm_capture_queue_buffer_.resize(aecm_render_queue_element_max_size_);
|
2016-10-22 05:04:30 -07:00
|
|
|
} else {
|
2016-10-25 05:42:20 -07:00
|
|
|
aecm_render_signal_queue_->Clear();
|
2016-10-25 04:45:24 -07:00
|
|
|
}
|
|
|
|
|
|
2016-10-25 05:42:20 -07:00
|
|
|
if (agc_render_queue_element_max_size_ <
|
|
|
|
|
new_agc_render_queue_element_max_size) {
|
|
|
|
|
agc_render_queue_element_max_size_ = new_agc_render_queue_element_max_size;
|
2016-10-25 04:45:24 -07:00
|
|
|
|
|
|
|
|
std::vector<int16_t> template_queue_element(
|
2016-10-25 05:42:20 -07:00
|
|
|
agc_render_queue_element_max_size_);
|
2016-10-25 04:45:24 -07:00
|
|
|
|
2016-10-25 05:42:20 -07:00
|
|
|
agc_render_signal_queue_.reset(
|
2016-10-25 04:45:24 -07:00
|
|
|
new SwapQueue<std::vector<int16_t>, RenderQueueItemVerifier<int16_t>>(
|
|
|
|
|
kMaxNumFramesToBuffer, template_queue_element,
|
|
|
|
|
RenderQueueItemVerifier<int16_t>(
|
2016-10-25 05:42:20 -07:00
|
|
|
agc_render_queue_element_max_size_)));
|
2016-10-25 04:45:24 -07:00
|
|
|
|
2016-10-25 05:42:20 -07:00
|
|
|
agc_render_queue_buffer_.resize(agc_render_queue_element_max_size_);
|
|
|
|
|
agc_capture_queue_buffer_.resize(agc_render_queue_element_max_size_);
|
2016-10-25 04:45:24 -07:00
|
|
|
} else {
|
2016-10-25 05:42:20 -07:00
|
|
|
agc_render_signal_queue_->Clear();
|
2016-10-22 05:04:30 -07:00
|
|
|
}
|
2016-10-28 05:39:16 -07:00
|
|
|
|
|
|
|
|
if (red_render_queue_element_max_size_ <
|
|
|
|
|
new_red_render_queue_element_max_size) {
|
|
|
|
|
red_render_queue_element_max_size_ = new_red_render_queue_element_max_size;
|
|
|
|
|
|
|
|
|
|
std::vector<float> template_queue_element(
|
|
|
|
|
red_render_queue_element_max_size_);
|
|
|
|
|
|
|
|
|
|
red_render_signal_queue_.reset(
|
|
|
|
|
new SwapQueue<std::vector<float>, RenderQueueItemVerifier<float>>(
|
|
|
|
|
kMaxNumFramesToBuffer, template_queue_element,
|
|
|
|
|
RenderQueueItemVerifier<float>(
|
|
|
|
|
red_render_queue_element_max_size_)));
|
|
|
|
|
|
|
|
|
|
red_render_queue_buffer_.resize(red_render_queue_element_max_size_);
|
|
|
|
|
red_capture_queue_buffer_.resize(red_render_queue_element_max_size_);
|
|
|
|
|
} else {
|
|
|
|
|
red_render_signal_queue_->Clear();
|
|
|
|
|
}
|
2016-10-22 05:04:30 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void AudioProcessingImpl::EmptyQueuedRenderAudio() {
|
|
|
|
|
rtc::CritScope cs_capture(&crit_capture_);
|
2016-10-25 05:42:20 -07:00
|
|
|
while (aec_render_signal_queue_->Remove(&aec_capture_queue_buffer_)) {
|
2016-10-22 05:04:30 -07:00
|
|
|
public_submodules_->echo_cancellation->ProcessRenderAudio(
|
2016-10-25 05:42:20 -07:00
|
|
|
aec_capture_queue_buffer_);
|
2016-10-25 04:45:24 -07:00
|
|
|
}
|
|
|
|
|
|
2016-10-25 05:42:20 -07:00
|
|
|
while (aecm_render_signal_queue_->Remove(&aecm_capture_queue_buffer_)) {
|
2016-10-25 04:45:24 -07:00
|
|
|
public_submodules_->echo_control_mobile->ProcessRenderAudio(
|
2016-10-25 05:42:20 -07:00
|
|
|
aecm_capture_queue_buffer_);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
while (agc_render_signal_queue_->Remove(&agc_capture_queue_buffer_)) {
|
|
|
|
|
public_submodules_->gain_control->ProcessRenderAudio(
|
|
|
|
|
agc_capture_queue_buffer_);
|
2016-10-22 05:04:30 -07:00
|
|
|
}
|
2016-10-28 05:39:16 -07:00
|
|
|
|
|
|
|
|
while (red_render_signal_queue_->Remove(&red_capture_queue_buffer_)) {
|
|
|
|
|
private_submodules_->residual_echo_detector->AnalyzeRenderAudio(
|
|
|
|
|
red_capture_queue_buffer_);
|
|
|
|
|
}
|
2016-10-22 05:04:30 -07:00
|
|
|
}
|
|
|
|
|
|
2014-03-04 20:58:13 +00:00
|
|
|
int AudioProcessingImpl::ProcessStream(AudioFrame* frame) {
|
2015-12-17 06:42:29 -08:00
|
|
|
TRACE_EVENT0("webrtc", "AudioProcessing::ProcessStream_AudioFrame");
|
2015-11-28 12:35:15 -08:00
|
|
|
{
|
|
|
|
|
// Acquire the capture lock in order to safely call the function
|
|
|
|
|
// that retrieves the render side data. This function accesses apm
|
|
|
|
|
// getters that need the capture lock held when being called.
|
|
|
|
|
// The lock needs to be released as
|
|
|
|
|
// public_submodules_->echo_control_mobile->is_enabled() aquires this lock
|
|
|
|
|
// as well.
|
|
|
|
|
rtc::CritScope cs_capture(&crit_capture_);
|
2016-10-22 05:04:30 -07:00
|
|
|
EmptyQueuedRenderAudio();
|
2015-11-28 12:35:15 -08:00
|
|
|
}
|
2015-11-16 16:27:42 -08:00
|
|
|
|
2014-03-04 20:58:13 +00:00
|
|
|
if (!frame) {
|
2014-01-07 17:45:09 +00:00
|
|
|
return kNullPointerError;
|
|
|
|
|
}
|
2014-04-22 21:00:04 +00:00
|
|
|
// Must be a native rate.
|
|
|
|
|
if (frame->sample_rate_hz_ != kSampleRate8kHz &&
|
|
|
|
|
frame->sample_rate_hz_ != kSampleRate16kHz &&
|
2014-11-17 23:01:23 +00:00
|
|
|
frame->sample_rate_hz_ != kSampleRate32kHz &&
|
|
|
|
|
frame->sample_rate_hz_ != kSampleRate48kHz) {
|
2014-04-22 21:00:04 +00:00
|
|
|
return kBadSampleRateError;
|
|
|
|
|
}
|
2015-11-17 02:16:45 -08:00
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
ProcessingConfig processing_config;
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
bool reinitialization_required = false;
|
2015-11-28 12:35:15 -08:00
|
|
|
{
|
|
|
|
|
// Aquire lock for the access of api_format.
|
|
|
|
|
// The lock is released immediately due to the conditional
|
|
|
|
|
// reinitialization.
|
|
|
|
|
rtc::CritScope cs_capture(&crit_capture_);
|
|
|
|
|
// TODO(ajm): The input and output rates and channels are currently
|
|
|
|
|
// constrained to be identical in the int16 interface.
|
|
|
|
|
processing_config = formats_.api_format;
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
|
|
|
|
|
reinitialization_required = UpdateActiveSubmoduleStates();
|
2015-11-28 12:35:15 -08:00
|
|
|
}
|
2015-07-23 11:41:39 -07:00
|
|
|
processing_config.input_stream().set_sample_rate_hz(frame->sample_rate_hz_);
|
|
|
|
|
processing_config.input_stream().set_num_channels(frame->num_channels_);
|
|
|
|
|
processing_config.output_stream().set_sample_rate_hz(frame->sample_rate_hz_);
|
|
|
|
|
processing_config.output_stream().set_num_channels(frame->num_channels_);
|
|
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
{
|
|
|
|
|
// Do conditional reinitialization.
|
|
|
|
|
rtc::CritScope cs_render(&crit_render_);
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
RETURN_ON_ERR(
|
|
|
|
|
MaybeInitializeCapture(processing_config, reinitialization_required));
|
2015-11-28 12:35:15 -08:00
|
|
|
}
|
|
|
|
|
rtc::CritScope cs_capture(&crit_capture_);
|
2015-11-17 02:16:45 -08:00
|
|
|
if (frame->samples_per_channel_ !=
|
2015-11-28 12:35:15 -08:00
|
|
|
formats_.api_format.input_stream().num_frames()) {
|
2011-07-07 08:21:25 +00:00
|
|
|
return kBadDataLengthError;
|
|
|
|
|
}
|
|
|
|
|
|
2011-12-03 00:03:31 +00:00
|
|
|
#ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP
|
FileWrapper[Impl] modifications and actually remove the "Impl" class.
This is a somewhat involved refactoring of this class. Here's an overview of the changes:
* FileWrapper can now be used as a regular class and instances allocated on the stack.
* The type now has support for move semantics and copy isn't allowed.
* New public ctor with FILE* that can be used instead of OpenFromFileHandle.
* New static Open() method. The intent of this is to allow opening a file and getting back a FileWrapper instance. Using this method instead of Create(), will allow us in the future to make the FILE* member pointer, to be const and simplify threading (get rid of the lock).
* Rename the Open() method to is_open() and make it inline.
* The FileWrapper interface is no longer a pure virtual interface. There's only one implementation so there's no need to go through a vtable for everything.
* Functionality offered by the class, is now reduced. No support for looping (not clear if that was actually useful to users of that flag), no need to implement the 'read_only_' functionality in the class, since file APIs implement that already, no support for *not* managing the file handle (this wasn't used). OpenFromFileHandle always "manages" the file.
* Delete the unused WriteText() method and don't support opening files in text mode. Text mode is only different on Windows and on Windows it translates \n to \r\n, which means that files such as log files, could have a slightly different format on Windows than other platforms. Besides, tools on Windows can handle UNIX line endings.
* Remove FileName(), change Trace code to manage its own path.
* Rename id_ member variable to file_.
* Removed the open_ member variable since the same functionality can be gotten from just checking the file pointer.
* Don't call CloseFile inside of Write. Write shouldn't be changing the state of the class beyond just attempting to write.
* Remove concept of looping from FileWrapper and never close inside of Read()
* Changed stream base classes to inherit from a common base class instead of both defining the Rewind method. Ultimately, Id' like to remove these interfaces and just have FileWrapper.
* Remove read_only param from OpenFromFileHandle
* Renamed size_in_bytes_ to position_, since it gets set to 0 when Rewind() is called (and the size actually does not change).
* Switch out rw lock for CriticalSection. The r/w lock was only used for reading when checking the open_ flag.
BUG=
Review-Url: https://codereview.webrtc.org/2054373002
Cr-Commit-Position: refs/heads/master@{#13155}
2016-06-15 10:30:14 -07:00
|
|
|
if (debug_dump_.debug_file->is_open()) {
|
2016-08-18 06:48:33 -07:00
|
|
|
RETURN_ON_ERR(WriteConfigMessage(false));
|
|
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
debug_dump_.capture.event_msg->set_type(audioproc::Event::STREAM);
|
|
|
|
|
audioproc::Stream* msg = debug_dump_.capture.event_msg->mutable_stream();
|
2015-07-23 11:41:39 -07:00
|
|
|
const size_t data_size =
|
|
|
|
|
sizeof(int16_t) * frame->samples_per_channel_ * frame->num_channels_;
|
2012-05-02 23:56:37 +00:00
|
|
|
msg->set_input_data(frame->data_, data_size);
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
2011-12-03 00:03:31 +00:00
|
|
|
#endif
|
2011-07-07 08:21:25 +00:00
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
capture_.capture_audio->DeinterleaveFrom(frame);
|
2016-09-16 15:02:15 -07:00
|
|
|
RETURN_ON_ERR(ProcessCaptureStreamLocked());
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
capture_.capture_audio->InterleaveTo(
|
|
|
|
|
frame, submodule_states_.CaptureMultiBandProcessingActive());
|
2014-03-04 20:58:13 +00:00
|
|
|
|
|
|
|
|
#ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP
|
FileWrapper[Impl] modifications and actually remove the "Impl" class.
This is a somewhat involved refactoring of this class. Here's an overview of the changes:
* FileWrapper can now be used as a regular class and instances allocated on the stack.
* The type now has support for move semantics and copy isn't allowed.
* New public ctor with FILE* that can be used instead of OpenFromFileHandle.
* New static Open() method. The intent of this is to allow opening a file and getting back a FileWrapper instance. Using this method instead of Create(), will allow us in the future to make the FILE* member pointer, to be const and simplify threading (get rid of the lock).
* Rename the Open() method to is_open() and make it inline.
* The FileWrapper interface is no longer a pure virtual interface. There's only one implementation so there's no need to go through a vtable for everything.
* Functionality offered by the class, is now reduced. No support for looping (not clear if that was actually useful to users of that flag), no need to implement the 'read_only_' functionality in the class, since file APIs implement that already, no support for *not* managing the file handle (this wasn't used). OpenFromFileHandle always "manages" the file.
* Delete the unused WriteText() method and don't support opening files in text mode. Text mode is only different on Windows and on Windows it translates \n to \r\n, which means that files such as log files, could have a slightly different format on Windows than other platforms. Besides, tools on Windows can handle UNIX line endings.
* Remove FileName(), change Trace code to manage its own path.
* Rename id_ member variable to file_.
* Removed the open_ member variable since the same functionality can be gotten from just checking the file pointer.
* Don't call CloseFile inside of Write. Write shouldn't be changing the state of the class beyond just attempting to write.
* Remove concept of looping from FileWrapper and never close inside of Read()
* Changed stream base classes to inherit from a common base class instead of both defining the Rewind method. Ultimately, Id' like to remove these interfaces and just have FileWrapper.
* Remove read_only param from OpenFromFileHandle
* Renamed size_in_bytes_ to position_, since it gets set to 0 when Rewind() is called (and the size actually does not change).
* Switch out rw lock for CriticalSection. The r/w lock was only used for reading when checking the open_ flag.
BUG=
Review-Url: https://codereview.webrtc.org/2054373002
Cr-Commit-Position: refs/heads/master@{#13155}
2016-06-15 10:30:14 -07:00
|
|
|
if (debug_dump_.debug_file->is_open()) {
|
2015-11-28 12:35:15 -08:00
|
|
|
audioproc::Stream* msg = debug_dump_.capture.event_msg->mutable_stream();
|
2015-07-23 11:41:39 -07:00
|
|
|
const size_t data_size =
|
|
|
|
|
sizeof(int16_t) * frame->samples_per_channel_ * frame->num_channels_;
|
2014-03-04 20:58:13 +00:00
|
|
|
msg->set_output_data(frame->data_, data_size);
|
2015-11-28 12:35:15 -08:00
|
|
|
RETURN_ON_ERR(WriteMessageToDebugFile(debug_dump_.debug_file.get(),
|
2016-01-15 03:06:36 -08:00
|
|
|
&debug_dump_.num_bytes_left_for_log_,
|
2015-11-28 12:35:15 -08:00
|
|
|
&crit_debug_, &debug_dump_.capture));
|
2014-03-04 20:58:13 +00:00
|
|
|
}
|
|
|
|
|
#endif
|
|
|
|
|
|
|
|
|
|
return kNoError;
|
|
|
|
|
}
|
|
|
|
|
|
2016-09-16 15:02:15 -07:00
|
|
|
int AudioProcessingImpl::ProcessCaptureStreamLocked() {
|
2016-03-15 09:34:24 -07:00
|
|
|
// Ensure that not both the AEC and AECM are active at the same time.
|
|
|
|
|
// TODO(peah): Simplify once the public API Enable functions for these
|
|
|
|
|
// are moved to APM.
|
|
|
|
|
RTC_DCHECK(!(public_submodules_->echo_cancellation->is_enabled() &&
|
|
|
|
|
public_submodules_->echo_control_mobile->is_enabled()));
|
|
|
|
|
|
2014-03-04 20:58:13 +00:00
|
|
|
#ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP
|
FileWrapper[Impl] modifications and actually remove the "Impl" class.
This is a somewhat involved refactoring of this class. Here's an overview of the changes:
* FileWrapper can now be used as a regular class and instances allocated on the stack.
* The type now has support for move semantics and copy isn't allowed.
* New public ctor with FILE* that can be used instead of OpenFromFileHandle.
* New static Open() method. The intent of this is to allow opening a file and getting back a FileWrapper instance. Using this method instead of Create(), will allow us in the future to make the FILE* member pointer, to be const and simplify threading (get rid of the lock).
* Rename the Open() method to is_open() and make it inline.
* The FileWrapper interface is no longer a pure virtual interface. There's only one implementation so there's no need to go through a vtable for everything.
* Functionality offered by the class, is now reduced. No support for looping (not clear if that was actually useful to users of that flag), no need to implement the 'read_only_' functionality in the class, since file APIs implement that already, no support for *not* managing the file handle (this wasn't used). OpenFromFileHandle always "manages" the file.
* Delete the unused WriteText() method and don't support opening files in text mode. Text mode is only different on Windows and on Windows it translates \n to \r\n, which means that files such as log files, could have a slightly different format on Windows than other platforms. Besides, tools on Windows can handle UNIX line endings.
* Remove FileName(), change Trace code to manage its own path.
* Rename id_ member variable to file_.
* Removed the open_ member variable since the same functionality can be gotten from just checking the file pointer.
* Don't call CloseFile inside of Write. Write shouldn't be changing the state of the class beyond just attempting to write.
* Remove concept of looping from FileWrapper and never close inside of Read()
* Changed stream base classes to inherit from a common base class instead of both defining the Rewind method. Ultimately, Id' like to remove these interfaces and just have FileWrapper.
* Remove read_only param from OpenFromFileHandle
* Renamed size_in_bytes_ to position_, since it gets set to 0 when Rewind() is called (and the size actually does not change).
* Switch out rw lock for CriticalSection. The r/w lock was only used for reading when checking the open_ flag.
BUG=
Review-Url: https://codereview.webrtc.org/2054373002
Cr-Commit-Position: refs/heads/master@{#13155}
2016-06-15 10:30:14 -07:00
|
|
|
if (debug_dump_.debug_file->is_open()) {
|
2015-11-28 12:35:15 -08:00
|
|
|
audioproc::Stream* msg = debug_dump_.capture.event_msg->mutable_stream();
|
|
|
|
|
msg->set_delay(capture_nonlocked_.stream_delay_ms);
|
|
|
|
|
msg->set_drift(
|
|
|
|
|
public_submodules_->echo_cancellation->stream_drift_samples());
|
2015-02-06 19:44:21 +00:00
|
|
|
msg->set_level(gain_control()->stream_analog_level());
|
2015-11-28 12:35:15 -08:00
|
|
|
msg->set_keypress(capture_.key_pressed);
|
2014-03-04 20:58:13 +00:00
|
|
|
}
|
|
|
|
|
#endif
|
2011-07-07 08:21:25 +00:00
|
|
|
|
2015-06-29 14:57:29 +02:00
|
|
|
MaybeUpdateHistograms();
|
|
|
|
|
|
2016-09-16 15:02:15 -07:00
|
|
|
AudioBuffer* capture_buffer = capture_.capture_audio.get(); // For brevity.
|
2015-08-14 10:35:55 -07:00
|
|
|
|
2016-02-13 16:40:47 -08:00
|
|
|
if (constants_.use_experimental_agc &&
|
2015-11-28 12:35:15 -08:00
|
|
|
public_submodules_->gain_control->is_enabled()) {
|
|
|
|
|
private_submodules_->agc_manager->AnalyzePreProcess(
|
2016-09-16 15:02:15 -07:00
|
|
|
capture_buffer->channels()[0], capture_buffer->num_channels(),
|
|
|
|
|
capture_nonlocked_.capture_processing_format.num_frames());
|
2014-12-15 09:41:24 +00:00
|
|
|
}
|
|
|
|
|
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
if (submodule_states_.CaptureMultiBandSubModulesActive() &&
|
|
|
|
|
SampleRateSupportsMultiBand(
|
2016-09-16 15:02:15 -07:00
|
|
|
capture_nonlocked_.capture_processing_format.sample_rate_hz())) {
|
|
|
|
|
capture_buffer->SplitIntoFrequencyBands();
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
2016-01-11 20:32:29 -08:00
|
|
|
if (capture_nonlocked_.beamformer_enabled) {
|
2016-09-16 15:02:15 -07:00
|
|
|
private_submodules_->beamformer->AnalyzeChunk(
|
|
|
|
|
*capture_buffer->split_data_f());
|
2016-07-01 17:19:09 -07:00
|
|
|
// Discards all channels by the leftmost one.
|
2016-09-16 15:02:15 -07:00
|
|
|
capture_buffer->set_num_channels(1);
|
2014-12-19 19:57:34 +00:00
|
|
|
}
|
|
|
|
|
|
2016-11-22 07:24:52 -08:00
|
|
|
if (private_submodules_->low_cut_filter) {
|
|
|
|
|
private_submodules_->low_cut_filter->Process(capture_buffer);
|
|
|
|
|
}
|
2016-09-16 15:02:15 -07:00
|
|
|
RETURN_ON_ERR(
|
|
|
|
|
public_submodules_->gain_control->AnalyzeCaptureAudio(capture_buffer));
|
|
|
|
|
public_submodules_->noise_suppression->AnalyzeCaptureAudio(capture_buffer);
|
2016-03-15 09:34:24 -07:00
|
|
|
|
|
|
|
|
// Ensure that the stream delay was set before the call to the
|
|
|
|
|
// AEC ProcessCaptureAudio function.
|
|
|
|
|
if (public_submodules_->echo_cancellation->is_enabled() &&
|
|
|
|
|
!was_stream_delay_set()) {
|
|
|
|
|
return AudioProcessing::kStreamParameterNotSetError;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
RETURN_ON_ERR(public_submodules_->echo_cancellation->ProcessCaptureAudio(
|
2016-09-16 15:02:15 -07:00
|
|
|
capture_buffer, stream_delay_ms()));
|
2011-07-07 08:21:25 +00:00
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
if (public_submodules_->echo_control_mobile->is_enabled() &&
|
|
|
|
|
public_submodules_->noise_suppression->is_enabled()) {
|
2016-09-16 15:02:15 -07:00
|
|
|
capture_buffer->CopyLowPassToReference();
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
2016-09-16 15:02:15 -07:00
|
|
|
public_submodules_->noise_suppression->ProcessCaptureAudio(capture_buffer);
|
2016-08-26 07:16:04 -07:00
|
|
|
#if WEBRTC_INTELLIGIBILITY_ENHANCER
|
2016-05-16 15:32:38 -07:00
|
|
|
if (capture_nonlocked_.intelligibility_enabled) {
|
2016-02-10 12:03:00 -08:00
|
|
|
RTC_DCHECK(public_submodules_->noise_suppression->is_enabled());
|
2016-05-16 15:32:38 -07:00
|
|
|
int gain_db = public_submodules_->gain_control->is_enabled() ?
|
|
|
|
|
public_submodules_->gain_control->compression_gain_db() :
|
|
|
|
|
0;
|
2016-06-30 15:35:41 -07:00
|
|
|
float gain = std::pow(10.f, gain_db / 20.f);
|
|
|
|
|
gain *= capture_nonlocked_.level_controller_enabled ?
|
|
|
|
|
private_submodules_->level_controller->GetLastGain() :
|
|
|
|
|
1.f;
|
2016-02-10 12:03:00 -08:00
|
|
|
public_submodules_->intelligibility_enhancer->SetCaptureNoiseEstimate(
|
2016-06-30 15:35:41 -07:00
|
|
|
public_submodules_->noise_suppression->NoiseEstimate(), gain);
|
2016-02-10 12:03:00 -08:00
|
|
|
}
|
2016-08-26 07:16:04 -07:00
|
|
|
#endif
|
2016-03-15 04:32:28 -07:00
|
|
|
|
|
|
|
|
// Ensure that the stream delay was set before the call to the
|
|
|
|
|
// AECM ProcessCaptureAudio function.
|
|
|
|
|
if (public_submodules_->echo_control_mobile->is_enabled() &&
|
|
|
|
|
!was_stream_delay_set()) {
|
|
|
|
|
return AudioProcessing::kStreamParameterNotSetError;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
RETURN_ON_ERR(public_submodules_->echo_control_mobile->ProcessCaptureAudio(
|
2016-09-16 15:02:15 -07:00
|
|
|
capture_buffer, stream_delay_ms()));
|
2016-03-15 04:32:28 -07:00
|
|
|
|
2016-10-28 05:39:16 -07:00
|
|
|
if (config_.residual_echo_detector.enabled) {
|
|
|
|
|
private_submodules_->residual_echo_detector->AnalyzeCaptureAudio(
|
|
|
|
|
rtc::ArrayView<const float>(
|
|
|
|
|
capture_buffer->split_bands_const_f(0)[kBand0To8kHz],
|
|
|
|
|
capture_buffer->num_frames_per_band()));
|
|
|
|
|
}
|
|
|
|
|
|
2016-07-01 17:19:09 -07:00
|
|
|
if (capture_nonlocked_.beamformer_enabled) {
|
2016-09-16 15:02:15 -07:00
|
|
|
private_submodules_->beamformer->PostFilter(capture_buffer->split_data_f());
|
2016-07-01 17:19:09 -07:00
|
|
|
}
|
|
|
|
|
|
2016-09-16 15:02:15 -07:00
|
|
|
public_submodules_->voice_detection->ProcessCaptureAudio(capture_buffer);
|
2014-12-15 09:41:24 +00:00
|
|
|
|
2016-02-13 16:40:47 -08:00
|
|
|
if (constants_.use_experimental_agc &&
|
2015-11-28 12:35:15 -08:00
|
|
|
public_submodules_->gain_control->is_enabled() &&
|
2016-01-11 20:32:29 -08:00
|
|
|
(!capture_nonlocked_.beamformer_enabled ||
|
2015-11-28 12:35:15 -08:00
|
|
|
private_submodules_->beamformer->is_target_present())) {
|
|
|
|
|
private_submodules_->agc_manager->Process(
|
2016-09-16 15:02:15 -07:00
|
|
|
capture_buffer->split_bands_const(0)[kBand0To8kHz],
|
|
|
|
|
capture_buffer->num_frames_per_band(), capture_nonlocked_.split_rate);
|
2014-12-15 09:41:24 +00:00
|
|
|
}
|
2016-03-15 02:28:08 -07:00
|
|
|
RETURN_ON_ERR(public_submodules_->gain_control->ProcessCaptureAudio(
|
2016-09-16 15:02:15 -07:00
|
|
|
capture_buffer, echo_cancellation()->stream_has_echo()));
|
2011-07-07 08:21:25 +00:00
|
|
|
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
if (submodule_states_.CaptureMultiBandProcessingActive() &&
|
|
|
|
|
SampleRateSupportsMultiBand(
|
2016-09-16 15:02:15 -07:00
|
|
|
capture_nonlocked_.capture_processing_format.sample_rate_hz())) {
|
|
|
|
|
capture_buffer->MergeFrequencyBands();
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
2014-12-15 09:41:24 +00:00
|
|
|
// TODO(aluebs): Investigate if the transient suppression placement should be
|
|
|
|
|
// before or after the AGC.
|
2015-11-28 12:35:15 -08:00
|
|
|
if (capture_.transient_suppressor_enabled) {
|
2014-12-15 09:41:24 +00:00
|
|
|
float voice_probability =
|
2015-11-28 12:35:15 -08:00
|
|
|
private_submodules_->agc_manager.get()
|
|
|
|
|
? private_submodules_->agc_manager->voice_probability()
|
|
|
|
|
: 1.f;
|
2014-12-15 09:41:24 +00:00
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
public_submodules_->transient_suppressor->Suppress(
|
2016-09-16 15:02:15 -07:00
|
|
|
capture_buffer->channels_f()[0], capture_buffer->num_frames(),
|
|
|
|
|
capture_buffer->num_channels(),
|
|
|
|
|
capture_buffer->split_bands_const_f(0)[kBand0To8kHz],
|
|
|
|
|
capture_buffer->num_frames_per_band(), capture_buffer->keyboard_data(),
|
|
|
|
|
capture_buffer->num_keyboard_frames(), voice_probability,
|
2015-11-28 12:35:15 -08:00
|
|
|
capture_.key_pressed);
|
2014-12-15 09:41:24 +00:00
|
|
|
}
|
|
|
|
|
|
2016-06-29 15:26:12 -07:00
|
|
|
if (capture_nonlocked_.level_controller_enabled) {
|
2016-09-16 15:02:15 -07:00
|
|
|
private_submodules_->level_controller->Process(capture_buffer);
|
2016-06-29 15:26:12 -07:00
|
|
|
}
|
|
|
|
|
|
2011-11-15 16:57:56 +00:00
|
|
|
// The level estimator operates on the recombined data.
|
2016-09-16 15:02:15 -07:00
|
|
|
public_submodules_->level_estimator->ProcessStream(capture_buffer);
|
2014-03-04 20:58:13 +00:00
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
capture_.was_stream_delay_set = false;
|
2014-03-04 20:58:13 +00:00
|
|
|
return kNoError;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
int AudioProcessingImpl::AnalyzeReverseStream(const float* const* data,
|
Update a ton of audio code to use size_t more correctly and in general reduce
use of int16_t/uint16_t.
This is the upshot of a recommendation by henrik.lundin and kwiberg on an original small change ( https://webrtc-codereview.appspot.com/42569004/#ps1 ) to stop using int16_t just because values could fit in it, and is similar in nature to a previous "mass change to use size_t more" ( https://webrtc-codereview.appspot.com/23129004/ ) which also needed to be split up for review but to land all at once, since, like adding "const", such changes tend to cause a lot of transitive effects.
This was be reviewed and approved in pieces:
https://codereview.webrtc.org/1224093003
https://codereview.webrtc.org/1224123002
https://codereview.webrtc.org/1224163002
https://codereview.webrtc.org/1225133003
https://codereview.webrtc.org/1225173002
https://codereview.webrtc.org/1227163003
https://codereview.webrtc.org/1227203003
https://codereview.webrtc.org/1227213002
https://codereview.webrtc.org/1227893002
https://codereview.webrtc.org/1228793004
https://codereview.webrtc.org/1228803003
https://codereview.webrtc.org/1228823002
https://codereview.webrtc.org/1228823003
https://codereview.webrtc.org/1228843002
https://codereview.webrtc.org/1230693002
https://codereview.webrtc.org/1231713002
The change is being landed as TBR to all the folks who reviewed the above.
BUG=chromium:81439
TEST=none
R=andrew@webrtc.org, pbos@webrtc.org
TBR=aluebs, andrew, asapersson, henrika, hlundin, jan.skoglund, kwiberg, minyue, pbos, pthatcher
Review URL: https://codereview.webrtc.org/1230503003 .
Cr-Commit-Position: refs/heads/master@{#9768}
2015-08-24 14:52:23 -07:00
|
|
|
size_t samples_per_channel,
|
2016-09-16 15:02:15 -07:00
|
|
|
int sample_rate_hz,
|
2014-03-04 20:58:13 +00:00
|
|
|
ChannelLayout layout) {
|
2015-12-17 06:42:29 -08:00
|
|
|
TRACE_EVENT0("webrtc", "AudioProcessing::AnalyzeReverseStream_ChannelLayout");
|
2015-11-28 12:35:15 -08:00
|
|
|
rtc::CritScope cs(&crit_render_);
|
2015-07-23 11:41:39 -07:00
|
|
|
const StreamConfig reverse_config = {
|
2016-09-16 15:02:15 -07:00
|
|
|
sample_rate_hz, ChannelsFromLayout(layout), LayoutHasKeyboard(layout),
|
2015-07-23 11:41:39 -07:00
|
|
|
};
|
|
|
|
|
if (samples_per_channel != reverse_config.num_frames()) {
|
|
|
|
|
return kBadDataLengthError;
|
|
|
|
|
}
|
2015-11-28 12:35:15 -08:00
|
|
|
return AnalyzeReverseStreamLocked(data, reverse_config, reverse_config);
|
2015-08-14 10:35:55 -07:00
|
|
|
}
|
|
|
|
|
|
2016-09-16 15:02:15 -07:00
|
|
|
int AudioProcessingImpl::ProcessReverseStream(const float* const* src,
|
|
|
|
|
const StreamConfig& input_config,
|
|
|
|
|
const StreamConfig& output_config,
|
|
|
|
|
float* const* dest) {
|
2015-12-17 06:42:29 -08:00
|
|
|
TRACE_EVENT0("webrtc", "AudioProcessing::ProcessReverseStream_StreamConfig");
|
2015-11-28 12:35:15 -08:00
|
|
|
rtc::CritScope cs(&crit_render_);
|
2016-09-16 15:02:15 -07:00
|
|
|
RETURN_ON_ERR(AnalyzeReverseStreamLocked(src, input_config, output_config));
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
if (submodule_states_.RenderMultiBandProcessingActive()) {
|
2015-11-28 12:35:15 -08:00
|
|
|
render_.render_audio->CopyTo(formats_.api_format.reverse_output_stream(),
|
|
|
|
|
dest);
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
} else if (formats_.api_format.reverse_input_stream() !=
|
|
|
|
|
formats_.api_format.reverse_output_stream()) {
|
2016-09-16 15:02:15 -07:00
|
|
|
render_.render_converter->Convert(src, input_config.num_samples(), dest,
|
|
|
|
|
output_config.num_samples());
|
2015-08-14 10:35:55 -07:00
|
|
|
} else {
|
2016-09-16 15:02:15 -07:00
|
|
|
CopyAudioIfNeeded(src, input_config.num_frames(),
|
|
|
|
|
input_config.num_channels(), dest);
|
2015-08-14 10:35:55 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return kNoError;
|
2015-07-23 11:41:39 -07:00
|
|
|
}
|
|
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
int AudioProcessingImpl::AnalyzeReverseStreamLocked(
|
2015-08-14 10:35:55 -07:00
|
|
|
const float* const* src,
|
2016-09-16 15:02:15 -07:00
|
|
|
const StreamConfig& input_config,
|
|
|
|
|
const StreamConfig& output_config) {
|
2015-11-28 12:35:15 -08:00
|
|
|
if (src == nullptr) {
|
2014-03-04 20:58:13 +00:00
|
|
|
return kNullPointerError;
|
|
|
|
|
}
|
2011-11-15 16:57:56 +00:00
|
|
|
|
2016-09-16 15:02:15 -07:00
|
|
|
if (input_config.num_channels() == 0) {
|
2015-07-23 11:41:39 -07:00
|
|
|
return kBadNumberChannelsError;
|
2014-03-04 20:58:13 +00:00
|
|
|
}
|
2011-07-07 08:21:25 +00:00
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
ProcessingConfig processing_config = formats_.api_format;
|
2016-09-16 15:02:15 -07:00
|
|
|
processing_config.reverse_input_stream() = input_config;
|
|
|
|
|
processing_config.reverse_output_stream() = output_config;
|
2015-07-23 11:41:39 -07:00
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
RETURN_ON_ERR(MaybeInitializeRender(processing_config));
|
2016-09-16 15:02:15 -07:00
|
|
|
assert(input_config.num_frames() ==
|
|
|
|
|
formats_.api_format.reverse_input_stream().num_frames());
|
2015-07-23 11:41:39 -07:00
|
|
|
|
2011-12-03 00:03:31 +00:00
|
|
|
#ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP
|
FileWrapper[Impl] modifications and actually remove the "Impl" class.
This is a somewhat involved refactoring of this class. Here's an overview of the changes:
* FileWrapper can now be used as a regular class and instances allocated on the stack.
* The type now has support for move semantics and copy isn't allowed.
* New public ctor with FILE* that can be used instead of OpenFromFileHandle.
* New static Open() method. The intent of this is to allow opening a file and getting back a FileWrapper instance. Using this method instead of Create(), will allow us in the future to make the FILE* member pointer, to be const and simplify threading (get rid of the lock).
* Rename the Open() method to is_open() and make it inline.
* The FileWrapper interface is no longer a pure virtual interface. There's only one implementation so there's no need to go through a vtable for everything.
* Functionality offered by the class, is now reduced. No support for looping (not clear if that was actually useful to users of that flag), no need to implement the 'read_only_' functionality in the class, since file APIs implement that already, no support for *not* managing the file handle (this wasn't used). OpenFromFileHandle always "manages" the file.
* Delete the unused WriteText() method and don't support opening files in text mode. Text mode is only different on Windows and on Windows it translates \n to \r\n, which means that files such as log files, could have a slightly different format on Windows than other platforms. Besides, tools on Windows can handle UNIX line endings.
* Remove FileName(), change Trace code to manage its own path.
* Rename id_ member variable to file_.
* Removed the open_ member variable since the same functionality can be gotten from just checking the file pointer.
* Don't call CloseFile inside of Write. Write shouldn't be changing the state of the class beyond just attempting to write.
* Remove concept of looping from FileWrapper and never close inside of Read()
* Changed stream base classes to inherit from a common base class instead of both defining the Rewind method. Ultimately, Id' like to remove these interfaces and just have FileWrapper.
* Remove read_only param from OpenFromFileHandle
* Renamed size_in_bytes_ to position_, since it gets set to 0 when Rewind() is called (and the size actually does not change).
* Switch out rw lock for CriticalSection. The r/w lock was only used for reading when checking the open_ flag.
BUG=
Review-Url: https://codereview.webrtc.org/2054373002
Cr-Commit-Position: refs/heads/master@{#13155}
2016-06-15 10:30:14 -07:00
|
|
|
if (debug_dump_.debug_file->is_open()) {
|
2015-11-28 12:35:15 -08:00
|
|
|
debug_dump_.render.event_msg->set_type(audioproc::Event::REVERSE_STREAM);
|
|
|
|
|
audioproc::ReverseStream* msg =
|
|
|
|
|
debug_dump_.render.event_msg->mutable_reverse_stream();
|
2014-08-28 10:43:09 +00:00
|
|
|
const size_t channel_size =
|
2015-11-28 12:35:15 -08:00
|
|
|
sizeof(float) * formats_.api_format.reverse_input_stream().num_frames();
|
Convert channel counts to size_t.
IIRC, this was originally requested by ajm during review of the other size_t conversions I did over the past year, and I agreed it made sense, but wanted to do it separately since those changes were already gargantuan.
BUG=chromium:81439
TEST=none
R=henrik.lundin@webrtc.org, henrika@webrtc.org, kjellander@webrtc.org, minyue@webrtc.org, perkj@webrtc.org, solenberg@webrtc.org, stefan@webrtc.org, tina.legrand@webrtc.org
Review URL: https://codereview.webrtc.org/1316523002 .
Cr-Commit-Position: refs/heads/master@{#11229}
2016-01-12 16:26:35 -08:00
|
|
|
for (size_t i = 0;
|
2015-11-28 12:35:15 -08:00
|
|
|
i < formats_.api_format.reverse_input_stream().num_channels(); ++i)
|
2015-08-14 10:35:55 -07:00
|
|
|
msg->add_channel(src[i], channel_size);
|
2015-11-28 12:35:15 -08:00
|
|
|
RETURN_ON_ERR(WriteMessageToDebugFile(debug_dump_.debug_file.get(),
|
2016-01-15 03:06:36 -08:00
|
|
|
&debug_dump_.num_bytes_left_for_log_,
|
2015-11-28 12:35:15 -08:00
|
|
|
&crit_debug_, &debug_dump_.render));
|
2011-08-03 21:08:51 +00:00
|
|
|
}
|
2011-12-03 00:03:31 +00:00
|
|
|
#endif
|
2011-08-03 21:08:51 +00:00
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
render_.render_audio->CopyFrom(src,
|
|
|
|
|
formats_.api_format.reverse_input_stream());
|
2016-09-16 15:02:15 -07:00
|
|
|
return ProcessRenderStreamLocked();
|
2015-08-14 10:35:55 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
int AudioProcessingImpl::ProcessReverseStream(AudioFrame* frame) {
|
2015-12-17 06:42:29 -08:00
|
|
|
TRACE_EVENT0("webrtc", "AudioProcessing::ProcessReverseStream_AudioFrame");
|
2015-11-28 12:35:15 -08:00
|
|
|
rtc::CritScope cs(&crit_render_);
|
|
|
|
|
if (frame == nullptr) {
|
2011-07-07 08:21:25 +00:00
|
|
|
return kNullPointerError;
|
|
|
|
|
}
|
2014-04-22 21:00:04 +00:00
|
|
|
// Must be a native rate.
|
|
|
|
|
if (frame->sample_rate_hz_ != kSampleRate8kHz &&
|
|
|
|
|
frame->sample_rate_hz_ != kSampleRate16kHz &&
|
2014-11-17 23:01:23 +00:00
|
|
|
frame->sample_rate_hz_ != kSampleRate32kHz &&
|
|
|
|
|
frame->sample_rate_hz_ != kSampleRate48kHz) {
|
2014-04-22 21:00:04 +00:00
|
|
|
return kBadSampleRateError;
|
|
|
|
|
}
|
2014-03-10 22:26:12 +00:00
|
|
|
|
2015-07-23 11:41:39 -07:00
|
|
|
if (frame->num_channels_ <= 0) {
|
|
|
|
|
return kBadNumberChannelsError;
|
|
|
|
|
}
|
|
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
ProcessingConfig processing_config = formats_.api_format;
|
2015-08-14 10:35:55 -07:00
|
|
|
processing_config.reverse_input_stream().set_sample_rate_hz(
|
|
|
|
|
frame->sample_rate_hz_);
|
|
|
|
|
processing_config.reverse_input_stream().set_num_channels(
|
|
|
|
|
frame->num_channels_);
|
|
|
|
|
processing_config.reverse_output_stream().set_sample_rate_hz(
|
|
|
|
|
frame->sample_rate_hz_);
|
|
|
|
|
processing_config.reverse_output_stream().set_num_channels(
|
|
|
|
|
frame->num_channels_);
|
2015-07-23 11:41:39 -07:00
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
RETURN_ON_ERR(MaybeInitializeRender(processing_config));
|
2015-07-23 11:41:39 -07:00
|
|
|
if (frame->samples_per_channel_ !=
|
2015-11-28 12:35:15 -08:00
|
|
|
formats_.api_format.reverse_input_stream().num_frames()) {
|
2014-03-04 20:58:13 +00:00
|
|
|
return kBadDataLengthError;
|
|
|
|
|
}
|
2011-07-07 08:21:25 +00:00
|
|
|
|
2011-12-03 00:03:31 +00:00
|
|
|
#ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP
|
FileWrapper[Impl] modifications and actually remove the "Impl" class.
This is a somewhat involved refactoring of this class. Here's an overview of the changes:
* FileWrapper can now be used as a regular class and instances allocated on the stack.
* The type now has support for move semantics and copy isn't allowed.
* New public ctor with FILE* that can be used instead of OpenFromFileHandle.
* New static Open() method. The intent of this is to allow opening a file and getting back a FileWrapper instance. Using this method instead of Create(), will allow us in the future to make the FILE* member pointer, to be const and simplify threading (get rid of the lock).
* Rename the Open() method to is_open() and make it inline.
* The FileWrapper interface is no longer a pure virtual interface. There's only one implementation so there's no need to go through a vtable for everything.
* Functionality offered by the class, is now reduced. No support for looping (not clear if that was actually useful to users of that flag), no need to implement the 'read_only_' functionality in the class, since file APIs implement that already, no support for *not* managing the file handle (this wasn't used). OpenFromFileHandle always "manages" the file.
* Delete the unused WriteText() method and don't support opening files in text mode. Text mode is only different on Windows and on Windows it translates \n to \r\n, which means that files such as log files, could have a slightly different format on Windows than other platforms. Besides, tools on Windows can handle UNIX line endings.
* Remove FileName(), change Trace code to manage its own path.
* Rename id_ member variable to file_.
* Removed the open_ member variable since the same functionality can be gotten from just checking the file pointer.
* Don't call CloseFile inside of Write. Write shouldn't be changing the state of the class beyond just attempting to write.
* Remove concept of looping from FileWrapper and never close inside of Read()
* Changed stream base classes to inherit from a common base class instead of both defining the Rewind method. Ultimately, Id' like to remove these interfaces and just have FileWrapper.
* Remove read_only param from OpenFromFileHandle
* Renamed size_in_bytes_ to position_, since it gets set to 0 when Rewind() is called (and the size actually does not change).
* Switch out rw lock for CriticalSection. The r/w lock was only used for reading when checking the open_ flag.
BUG=
Review-Url: https://codereview.webrtc.org/2054373002
Cr-Commit-Position: refs/heads/master@{#13155}
2016-06-15 10:30:14 -07:00
|
|
|
if (debug_dump_.debug_file->is_open()) {
|
2015-11-28 12:35:15 -08:00
|
|
|
debug_dump_.render.event_msg->set_type(audioproc::Event::REVERSE_STREAM);
|
|
|
|
|
audioproc::ReverseStream* msg =
|
|
|
|
|
debug_dump_.render.event_msg->mutable_reverse_stream();
|
2015-07-23 11:41:39 -07:00
|
|
|
const size_t data_size =
|
|
|
|
|
sizeof(int16_t) * frame->samples_per_channel_ * frame->num_channels_;
|
2012-05-02 23:56:37 +00:00
|
|
|
msg->set_data(frame->data_, data_size);
|
2015-11-28 12:35:15 -08:00
|
|
|
RETURN_ON_ERR(WriteMessageToDebugFile(debug_dump_.debug_file.get(),
|
2016-01-15 03:06:36 -08:00
|
|
|
&debug_dump_.num_bytes_left_for_log_,
|
2015-11-28 12:35:15 -08:00
|
|
|
&crit_debug_, &debug_dump_.render));
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
2011-12-03 00:03:31 +00:00
|
|
|
#endif
|
2015-11-28 12:35:15 -08:00
|
|
|
render_.render_audio->DeinterleaveFrom(frame);
|
2016-09-16 15:02:15 -07:00
|
|
|
RETURN_ON_ERR(ProcessRenderStreamLocked());
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
render_.render_audio->InterleaveTo(
|
|
|
|
|
frame, submodule_states_.RenderMultiBandProcessingActive());
|
2016-03-17 20:39:53 -07:00
|
|
|
return kNoError;
|
2014-03-04 20:58:13 +00:00
|
|
|
}
|
2011-07-07 08:21:25 +00:00
|
|
|
|
2016-09-16 15:02:15 -07:00
|
|
|
int AudioProcessingImpl::ProcessRenderStreamLocked() {
|
|
|
|
|
AudioBuffer* render_buffer = render_.render_audio.get(); // For brevity.
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
if (submodule_states_.RenderMultiBandSubModulesActive() &&
|
2016-09-16 15:02:15 -07:00
|
|
|
SampleRateSupportsMultiBand(
|
|
|
|
|
formats_.render_processing_format.sample_rate_hz())) {
|
|
|
|
|
render_buffer->SplitIntoFrequencyBands();
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
2016-08-26 07:16:04 -07:00
|
|
|
#if WEBRTC_INTELLIGIBILITY_ENHANCER
|
2016-05-16 15:32:38 -07:00
|
|
|
if (capture_nonlocked_.intelligibility_enabled) {
|
2015-11-28 12:35:15 -08:00
|
|
|
public_submodules_->intelligibility_enhancer->ProcessRenderAudio(
|
2016-09-20 14:51:56 -07:00
|
|
|
render_buffer);
|
2015-08-14 10:35:55 -07:00
|
|
|
}
|
2016-08-26 07:16:04 -07:00
|
|
|
#endif
|
2015-08-14 10:35:55 -07:00
|
|
|
|
2016-10-22 05:04:30 -07:00
|
|
|
QueueRenderAudio(render_buffer);
|
2011-07-07 08:21:25 +00:00
|
|
|
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
if (submodule_states_.RenderMultiBandProcessingActive() &&
|
2016-09-16 15:02:15 -07:00
|
|
|
SampleRateSupportsMultiBand(
|
|
|
|
|
formats_.render_processing_format.sample_rate_hz())) {
|
|
|
|
|
render_buffer->MergeFrequencyBands();
|
2015-08-14 10:35:55 -07:00
|
|
|
}
|
|
|
|
|
|
2014-03-04 20:58:13 +00:00
|
|
|
return kNoError;
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
int AudioProcessingImpl::set_stream_delay_ms(int delay) {
|
2015-11-28 12:35:15 -08:00
|
|
|
rtc::CritScope cs(&crit_capture_);
|
2012-05-29 21:14:06 +00:00
|
|
|
Error retval = kNoError;
|
2015-11-28 12:35:15 -08:00
|
|
|
capture_.was_stream_delay_set = true;
|
|
|
|
|
delay += capture_.delay_offset_ms;
|
2012-03-06 19:03:39 +00:00
|
|
|
|
2011-07-07 08:21:25 +00:00
|
|
|
if (delay < 0) {
|
2012-05-29 21:14:06 +00:00
|
|
|
delay = 0;
|
|
|
|
|
retval = kBadStreamParameterWarning;
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// TODO(ajm): the max is rather arbitrarily chosen; investigate.
|
|
|
|
|
if (delay > 500) {
|
2012-05-29 21:14:06 +00:00
|
|
|
delay = 500;
|
|
|
|
|
retval = kBadStreamParameterWarning;
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
capture_nonlocked_.stream_delay_ms = delay;
|
2012-05-29 21:14:06 +00:00
|
|
|
return retval;
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
int AudioProcessingImpl::stream_delay_ms() const {
|
2015-11-28 12:35:15 -08:00
|
|
|
// Used as callback from submodules, hence locking is not allowed.
|
|
|
|
|
return capture_nonlocked_.stream_delay_ms;
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
bool AudioProcessingImpl::was_stream_delay_set() const {
|
2015-11-28 12:35:15 -08:00
|
|
|
// Used as callback from submodules, hence locking is not allowed.
|
|
|
|
|
return capture_.was_stream_delay_set;
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
2014-03-04 20:58:13 +00:00
|
|
|
void AudioProcessingImpl::set_stream_key_pressed(bool key_pressed) {
|
2015-11-28 12:35:15 -08:00
|
|
|
rtc::CritScope cs(&crit_capture_);
|
|
|
|
|
capture_.key_pressed = key_pressed;
|
2014-03-04 20:58:13 +00:00
|
|
|
}
|
|
|
|
|
|
2012-03-06 19:03:39 +00:00
|
|
|
void AudioProcessingImpl::set_delay_offset_ms(int offset) {
|
2015-11-28 12:35:15 -08:00
|
|
|
rtc::CritScope cs(&crit_capture_);
|
|
|
|
|
capture_.delay_offset_ms = offset;
|
2012-03-06 19:03:39 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
int AudioProcessingImpl::delay_offset_ms() const {
|
2015-11-28 12:35:15 -08:00
|
|
|
rtc::CritScope cs(&crit_capture_);
|
|
|
|
|
return capture_.delay_offset_ms;
|
2012-03-06 19:03:39 +00:00
|
|
|
}
|
|
|
|
|
|
2011-07-07 08:21:25 +00:00
|
|
|
int AudioProcessingImpl::StartDebugRecording(
|
2016-01-15 03:06:36 -08:00
|
|
|
const char filename[AudioProcessing::kMaxFilenameSize],
|
|
|
|
|
int64_t max_log_size_bytes) {
|
2015-11-28 12:35:15 -08:00
|
|
|
// Run in a single-threaded manner.
|
|
|
|
|
rtc::CritScope cs_render(&crit_render_);
|
|
|
|
|
rtc::CritScope cs_capture(&crit_capture_);
|
2015-05-20 11:11:07 +02:00
|
|
|
static_assert(kMaxFilenameSize == FileWrapper::kMaxFileNameSize, "");
|
2011-07-07 08:21:25 +00:00
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
if (filename == nullptr) {
|
2011-07-07 08:21:25 +00:00
|
|
|
return kNullPointerError;
|
|
|
|
|
}
|
|
|
|
|
|
2011-12-03 00:03:31 +00:00
|
|
|
#ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP
|
2016-01-15 03:06:36 -08:00
|
|
|
debug_dump_.num_bytes_left_for_log_ = max_log_size_bytes;
|
2011-07-07 08:21:25 +00:00
|
|
|
// Stop any ongoing recording.
|
FileWrapper[Impl] modifications and actually remove the "Impl" class.
This is a somewhat involved refactoring of this class. Here's an overview of the changes:
* FileWrapper can now be used as a regular class and instances allocated on the stack.
* The type now has support for move semantics and copy isn't allowed.
* New public ctor with FILE* that can be used instead of OpenFromFileHandle.
* New static Open() method. The intent of this is to allow opening a file and getting back a FileWrapper instance. Using this method instead of Create(), will allow us in the future to make the FILE* member pointer, to be const and simplify threading (get rid of the lock).
* Rename the Open() method to is_open() and make it inline.
* The FileWrapper interface is no longer a pure virtual interface. There's only one implementation so there's no need to go through a vtable for everything.
* Functionality offered by the class, is now reduced. No support for looping (not clear if that was actually useful to users of that flag), no need to implement the 'read_only_' functionality in the class, since file APIs implement that already, no support for *not* managing the file handle (this wasn't used). OpenFromFileHandle always "manages" the file.
* Delete the unused WriteText() method and don't support opening files in text mode. Text mode is only different on Windows and on Windows it translates \n to \r\n, which means that files such as log files, could have a slightly different format on Windows than other platforms. Besides, tools on Windows can handle UNIX line endings.
* Remove FileName(), change Trace code to manage its own path.
* Rename id_ member variable to file_.
* Removed the open_ member variable since the same functionality can be gotten from just checking the file pointer.
* Don't call CloseFile inside of Write. Write shouldn't be changing the state of the class beyond just attempting to write.
* Remove concept of looping from FileWrapper and never close inside of Read()
* Changed stream base classes to inherit from a common base class instead of both defining the Rewind method. Ultimately, Id' like to remove these interfaces and just have FileWrapper.
* Remove read_only param from OpenFromFileHandle
* Renamed size_in_bytes_ to position_, since it gets set to 0 when Rewind() is called (and the size actually does not change).
* Switch out rw lock for CriticalSection. The r/w lock was only used for reading when checking the open_ flag.
BUG=
Review-Url: https://codereview.webrtc.org/2054373002
Cr-Commit-Position: refs/heads/master@{#13155}
2016-06-15 10:30:14 -07:00
|
|
|
debug_dump_.debug_file->CloseFile();
|
2011-07-07 08:21:25 +00:00
|
|
|
|
FileWrapper[Impl] modifications and actually remove the "Impl" class.
This is a somewhat involved refactoring of this class. Here's an overview of the changes:
* FileWrapper can now be used as a regular class and instances allocated on the stack.
* The type now has support for move semantics and copy isn't allowed.
* New public ctor with FILE* that can be used instead of OpenFromFileHandle.
* New static Open() method. The intent of this is to allow opening a file and getting back a FileWrapper instance. Using this method instead of Create(), will allow us in the future to make the FILE* member pointer, to be const and simplify threading (get rid of the lock).
* Rename the Open() method to is_open() and make it inline.
* The FileWrapper interface is no longer a pure virtual interface. There's only one implementation so there's no need to go through a vtable for everything.
* Functionality offered by the class, is now reduced. No support for looping (not clear if that was actually useful to users of that flag), no need to implement the 'read_only_' functionality in the class, since file APIs implement that already, no support for *not* managing the file handle (this wasn't used). OpenFromFileHandle always "manages" the file.
* Delete the unused WriteText() method and don't support opening files in text mode. Text mode is only different on Windows and on Windows it translates \n to \r\n, which means that files such as log files, could have a slightly different format on Windows than other platforms. Besides, tools on Windows can handle UNIX line endings.
* Remove FileName(), change Trace code to manage its own path.
* Rename id_ member variable to file_.
* Removed the open_ member variable since the same functionality can be gotten from just checking the file pointer.
* Don't call CloseFile inside of Write. Write shouldn't be changing the state of the class beyond just attempting to write.
* Remove concept of looping from FileWrapper and never close inside of Read()
* Changed stream base classes to inherit from a common base class instead of both defining the Rewind method. Ultimately, Id' like to remove these interfaces and just have FileWrapper.
* Remove read_only param from OpenFromFileHandle
* Renamed size_in_bytes_ to position_, since it gets set to 0 when Rewind() is called (and the size actually does not change).
* Switch out rw lock for CriticalSection. The r/w lock was only used for reading when checking the open_ flag.
BUG=
Review-Url: https://codereview.webrtc.org/2054373002
Cr-Commit-Position: refs/heads/master@{#13155}
2016-06-15 10:30:14 -07:00
|
|
|
if (!debug_dump_.debug_file->OpenFile(filename, false)) {
|
2011-07-07 08:21:25 +00:00
|
|
|
return kFileError;
|
|
|
|
|
}
|
|
|
|
|
|
2015-10-03 00:39:14 +02:00
|
|
|
RETURN_ON_ERR(WriteConfigMessage(true));
|
|
|
|
|
RETURN_ON_ERR(WriteInitMessage());
|
2013-12-06 16:05:17 +00:00
|
|
|
return kNoError;
|
|
|
|
|
#else
|
|
|
|
|
return kUnsupportedFunctionError;
|
|
|
|
|
#endif // WEBRTC_AUDIOPROC_DEBUG_DUMP
|
|
|
|
|
}
|
|
|
|
|
|
2016-01-15 03:06:36 -08:00
|
|
|
int AudioProcessingImpl::StartDebugRecording(FILE* handle,
|
|
|
|
|
int64_t max_log_size_bytes) {
|
2015-11-28 12:35:15 -08:00
|
|
|
// Run in a single-threaded manner.
|
|
|
|
|
rtc::CritScope cs_render(&crit_render_);
|
|
|
|
|
rtc::CritScope cs_capture(&crit_capture_);
|
2013-12-06 16:05:17 +00:00
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
if (handle == nullptr) {
|
2013-12-06 16:05:17 +00:00
|
|
|
return kNullPointerError;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
#ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP
|
2016-01-15 03:06:36 -08:00
|
|
|
debug_dump_.num_bytes_left_for_log_ = max_log_size_bytes;
|
|
|
|
|
|
2013-12-06 16:05:17 +00:00
|
|
|
// Stop any ongoing recording.
|
FileWrapper[Impl] modifications and actually remove the "Impl" class.
This is a somewhat involved refactoring of this class. Here's an overview of the changes:
* FileWrapper can now be used as a regular class and instances allocated on the stack.
* The type now has support for move semantics and copy isn't allowed.
* New public ctor with FILE* that can be used instead of OpenFromFileHandle.
* New static Open() method. The intent of this is to allow opening a file and getting back a FileWrapper instance. Using this method instead of Create(), will allow us in the future to make the FILE* member pointer, to be const and simplify threading (get rid of the lock).
* Rename the Open() method to is_open() and make it inline.
* The FileWrapper interface is no longer a pure virtual interface. There's only one implementation so there's no need to go through a vtable for everything.
* Functionality offered by the class, is now reduced. No support for looping (not clear if that was actually useful to users of that flag), no need to implement the 'read_only_' functionality in the class, since file APIs implement that already, no support for *not* managing the file handle (this wasn't used). OpenFromFileHandle always "manages" the file.
* Delete the unused WriteText() method and don't support opening files in text mode. Text mode is only different on Windows and on Windows it translates \n to \r\n, which means that files such as log files, could have a slightly different format on Windows than other platforms. Besides, tools on Windows can handle UNIX line endings.
* Remove FileName(), change Trace code to manage its own path.
* Rename id_ member variable to file_.
* Removed the open_ member variable since the same functionality can be gotten from just checking the file pointer.
* Don't call CloseFile inside of Write. Write shouldn't be changing the state of the class beyond just attempting to write.
* Remove concept of looping from FileWrapper and never close inside of Read()
* Changed stream base classes to inherit from a common base class instead of both defining the Rewind method. Ultimately, Id' like to remove these interfaces and just have FileWrapper.
* Remove read_only param from OpenFromFileHandle
* Renamed size_in_bytes_ to position_, since it gets set to 0 when Rewind() is called (and the size actually does not change).
* Switch out rw lock for CriticalSection. The r/w lock was only used for reading when checking the open_ flag.
BUG=
Review-Url: https://codereview.webrtc.org/2054373002
Cr-Commit-Position: refs/heads/master@{#13155}
2016-06-15 10:30:14 -07:00
|
|
|
debug_dump_.debug_file->CloseFile();
|
2013-12-06 16:05:17 +00:00
|
|
|
|
FileWrapper[Impl] modifications and actually remove the "Impl" class.
This is a somewhat involved refactoring of this class. Here's an overview of the changes:
* FileWrapper can now be used as a regular class and instances allocated on the stack.
* The type now has support for move semantics and copy isn't allowed.
* New public ctor with FILE* that can be used instead of OpenFromFileHandle.
* New static Open() method. The intent of this is to allow opening a file and getting back a FileWrapper instance. Using this method instead of Create(), will allow us in the future to make the FILE* member pointer, to be const and simplify threading (get rid of the lock).
* Rename the Open() method to is_open() and make it inline.
* The FileWrapper interface is no longer a pure virtual interface. There's only one implementation so there's no need to go through a vtable for everything.
* Functionality offered by the class, is now reduced. No support for looping (not clear if that was actually useful to users of that flag), no need to implement the 'read_only_' functionality in the class, since file APIs implement that already, no support for *not* managing the file handle (this wasn't used). OpenFromFileHandle always "manages" the file.
* Delete the unused WriteText() method and don't support opening files in text mode. Text mode is only different on Windows and on Windows it translates \n to \r\n, which means that files such as log files, could have a slightly different format on Windows than other platforms. Besides, tools on Windows can handle UNIX line endings.
* Remove FileName(), change Trace code to manage its own path.
* Rename id_ member variable to file_.
* Removed the open_ member variable since the same functionality can be gotten from just checking the file pointer.
* Don't call CloseFile inside of Write. Write shouldn't be changing the state of the class beyond just attempting to write.
* Remove concept of looping from FileWrapper and never close inside of Read()
* Changed stream base classes to inherit from a common base class instead of both defining the Rewind method. Ultimately, Id' like to remove these interfaces and just have FileWrapper.
* Remove read_only param from OpenFromFileHandle
* Renamed size_in_bytes_ to position_, since it gets set to 0 when Rewind() is called (and the size actually does not change).
* Switch out rw lock for CriticalSection. The r/w lock was only used for reading when checking the open_ flag.
BUG=
Review-Url: https://codereview.webrtc.org/2054373002
Cr-Commit-Position: refs/heads/master@{#13155}
2016-06-15 10:30:14 -07:00
|
|
|
if (!debug_dump_.debug_file->OpenFromFileHandle(handle)) {
|
2013-12-06 16:05:17 +00:00
|
|
|
return kFileError;
|
|
|
|
|
}
|
|
|
|
|
|
2015-10-03 00:39:14 +02:00
|
|
|
RETURN_ON_ERR(WriteConfigMessage(true));
|
|
|
|
|
RETURN_ON_ERR(WriteInitMessage());
|
2011-07-07 08:21:25 +00:00
|
|
|
return kNoError;
|
2011-12-03 00:03:31 +00:00
|
|
|
#else
|
|
|
|
|
return kUnsupportedFunctionError;
|
|
|
|
|
#endif // WEBRTC_AUDIOPROC_DEBUG_DUMP
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
2016-10-12 03:01:49 -07:00
|
|
|
int AudioProcessingImpl::StartDebugRecording(FILE* handle) {
|
|
|
|
|
return StartDebugRecording(handle, -1);
|
|
|
|
|
}
|
|
|
|
|
|
2014-10-10 08:36:56 +00:00
|
|
|
int AudioProcessingImpl::StartDebugRecordingForPlatformFile(
|
|
|
|
|
rtc::PlatformFile handle) {
|
2015-11-28 12:35:15 -08:00
|
|
|
// Run in a single-threaded manner.
|
|
|
|
|
rtc::CritScope cs_render(&crit_render_);
|
|
|
|
|
rtc::CritScope cs_capture(&crit_capture_);
|
2014-10-10 08:36:56 +00:00
|
|
|
FILE* stream = rtc::FdopenPlatformFileForWriting(handle);
|
2016-01-15 03:06:36 -08:00
|
|
|
return StartDebugRecording(stream, -1);
|
2014-10-10 08:36:56 +00:00
|
|
|
}
|
|
|
|
|
|
2011-07-07 08:21:25 +00:00
|
|
|
int AudioProcessingImpl::StopDebugRecording() {
|
2015-11-28 12:35:15 -08:00
|
|
|
// Run in a single-threaded manner.
|
|
|
|
|
rtc::CritScope cs_render(&crit_render_);
|
|
|
|
|
rtc::CritScope cs_capture(&crit_capture_);
|
2011-12-03 00:03:31 +00:00
|
|
|
|
|
|
|
|
#ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP
|
2011-07-07 08:21:25 +00:00
|
|
|
// We just return if recording hasn't started.
|
FileWrapper[Impl] modifications and actually remove the "Impl" class.
This is a somewhat involved refactoring of this class. Here's an overview of the changes:
* FileWrapper can now be used as a regular class and instances allocated on the stack.
* The type now has support for move semantics and copy isn't allowed.
* New public ctor with FILE* that can be used instead of OpenFromFileHandle.
* New static Open() method. The intent of this is to allow opening a file and getting back a FileWrapper instance. Using this method instead of Create(), will allow us in the future to make the FILE* member pointer, to be const and simplify threading (get rid of the lock).
* Rename the Open() method to is_open() and make it inline.
* The FileWrapper interface is no longer a pure virtual interface. There's only one implementation so there's no need to go through a vtable for everything.
* Functionality offered by the class, is now reduced. No support for looping (not clear if that was actually useful to users of that flag), no need to implement the 'read_only_' functionality in the class, since file APIs implement that already, no support for *not* managing the file handle (this wasn't used). OpenFromFileHandle always "manages" the file.
* Delete the unused WriteText() method and don't support opening files in text mode. Text mode is only different on Windows and on Windows it translates \n to \r\n, which means that files such as log files, could have a slightly different format on Windows than other platforms. Besides, tools on Windows can handle UNIX line endings.
* Remove FileName(), change Trace code to manage its own path.
* Rename id_ member variable to file_.
* Removed the open_ member variable since the same functionality can be gotten from just checking the file pointer.
* Don't call CloseFile inside of Write. Write shouldn't be changing the state of the class beyond just attempting to write.
* Remove concept of looping from FileWrapper and never close inside of Read()
* Changed stream base classes to inherit from a common base class instead of both defining the Rewind method. Ultimately, Id' like to remove these interfaces and just have FileWrapper.
* Remove read_only param from OpenFromFileHandle
* Renamed size_in_bytes_ to position_, since it gets set to 0 when Rewind() is called (and the size actually does not change).
* Switch out rw lock for CriticalSection. The r/w lock was only used for reading when checking the open_ flag.
BUG=
Review-Url: https://codereview.webrtc.org/2054373002
Cr-Commit-Position: refs/heads/master@{#13155}
2016-06-15 10:30:14 -07:00
|
|
|
debug_dump_.debug_file->CloseFile();
|
2011-07-07 08:21:25 +00:00
|
|
|
return kNoError;
|
2011-12-03 00:03:31 +00:00
|
|
|
#else
|
|
|
|
|
return kUnsupportedFunctionError;
|
|
|
|
|
#endif // WEBRTC_AUDIOPROC_DEBUG_DUMP
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
2016-10-28 07:55:33 -07:00
|
|
|
// TODO(ivoc): Remove this when GetStatistics() becomes pure virtual.
|
|
|
|
|
AudioProcessing::AudioProcessingStatistics AudioProcessing::GetStatistics()
|
|
|
|
|
const {
|
|
|
|
|
return AudioProcessingStatistics();
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
AudioProcessing::AudioProcessingStatistics AudioProcessingImpl::GetStatistics()
|
|
|
|
|
const {
|
|
|
|
|
AudioProcessingStatistics stats;
|
|
|
|
|
EchoCancellation::Metrics metrics;
|
2016-11-02 09:14:37 -07:00
|
|
|
int success = public_submodules_->echo_cancellation->GetMetrics(&metrics);
|
|
|
|
|
if (success == Error::kNoError) {
|
|
|
|
|
stats.a_nlp.Set(metrics.a_nlp);
|
|
|
|
|
stats.divergent_filter_fraction = metrics.divergent_filter_fraction;
|
|
|
|
|
stats.echo_return_loss.Set(metrics.echo_return_loss);
|
|
|
|
|
stats.echo_return_loss_enhancement.Set(
|
|
|
|
|
metrics.echo_return_loss_enhancement);
|
|
|
|
|
stats.residual_echo_return_loss.Set(metrics.residual_echo_return_loss);
|
|
|
|
|
}
|
2016-11-14 07:55:03 -08:00
|
|
|
stats.residual_echo_likelihood =
|
|
|
|
|
private_submodules_->residual_echo_detector->echo_likelihood();
|
2016-10-28 07:55:33 -07:00
|
|
|
public_submodules_->echo_cancellation->GetDelayMetrics(
|
|
|
|
|
&stats.delay_median, &stats.delay_standard_deviation,
|
|
|
|
|
&stats.fraction_poor_delays);
|
|
|
|
|
return stats;
|
|
|
|
|
}
|
|
|
|
|
|
2011-07-07 08:21:25 +00:00
|
|
|
EchoCancellation* AudioProcessingImpl::echo_cancellation() const {
|
2016-03-05 03:01:14 -08:00
|
|
|
return public_submodules_->echo_cancellation.get();
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
EchoControlMobile* AudioProcessingImpl::echo_control_mobile() const {
|
2016-03-10 12:54:25 -08:00
|
|
|
return public_submodules_->echo_control_mobile.get();
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
GainControl* AudioProcessingImpl::gain_control() const {
|
2016-02-13 16:40:47 -08:00
|
|
|
if (constants_.use_experimental_agc) {
|
|
|
|
|
return public_submodules_->gain_control_for_experimental_agc.get();
|
2014-12-15 09:41:24 +00:00
|
|
|
}
|
2016-03-10 21:09:04 -08:00
|
|
|
return public_submodules_->gain_control.get();
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
HighPassFilter* AudioProcessingImpl::high_pass_filter() const {
|
2016-11-22 07:24:52 -08:00
|
|
|
return high_pass_filter_impl_.get();
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
LevelEstimator* AudioProcessingImpl::level_estimator() const {
|
2015-12-15 11:39:38 -08:00
|
|
|
return public_submodules_->level_estimator.get();
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
NoiseSuppression* AudioProcessingImpl::noise_suppression() const {
|
2015-12-08 13:22:33 -08:00
|
|
|
return public_submodules_->noise_suppression.get();
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
VoiceDetection* AudioProcessingImpl::voice_detection() const {
|
2015-12-16 03:31:12 -08:00
|
|
|
return public_submodules_->voice_detection.get();
|
2011-07-07 08:21:25 +00:00
|
|
|
}
|
|
|
|
|
|
2016-11-22 07:24:52 -08:00
|
|
|
void AudioProcessingImpl::MutateConfig(
|
|
|
|
|
rtc::FunctionView<void(AudioProcessing::Config*)> mutator) {
|
|
|
|
|
rtc::CritScope cs_render(&crit_render_);
|
|
|
|
|
rtc::CritScope cs_capture(&crit_capture_);
|
|
|
|
|
mutator(&config_);
|
|
|
|
|
ApplyConfig(config_);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
AudioProcessing::Config AudioProcessingImpl::GetConfig() const {
|
|
|
|
|
rtc::CritScope cs_render(&crit_render_);
|
|
|
|
|
rtc::CritScope cs_capture(&crit_capture_);
|
|
|
|
|
return config_;
|
|
|
|
|
}
|
|
|
|
|
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
bool AudioProcessingImpl::UpdateActiveSubmoduleStates() {
|
|
|
|
|
return submodule_states_.Update(
|
2016-11-22 07:24:52 -08:00
|
|
|
config_.high_pass_filter.enabled,
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
public_submodules_->echo_cancellation->is_enabled(),
|
|
|
|
|
public_submodules_->echo_control_mobile->is_enabled(),
|
2016-10-28 05:39:16 -07:00
|
|
|
config_.residual_echo_detector.enabled,
|
The audio processing module (APM) relies on two for
functionalities doing sample-rate conversions:
-The implicit resampling done in the AudioBuffer CopyTo,
CopyFrom, InterleaveTo and DeinterleaveFrom methods.
-The multi-band splitting scheme.
The selection of rates in these have been difficult and
complicated, partly due to that the APM API which allows
for activating the APM submodules without notifying
the APM.
This CL adds functionality that for each capture frame
polls all submodules for whether they are active or not
and compares this against a cached result.
Furthermore, new functionality is added that based on the
results of the comparison do a reinitialization of the APM.
This has several advantages
-The code deciding on whether to analysis and synthesis is
needed for the bandsplitting can be much simplified and
centralized.
-The selection of the processing rate can be done such as
to avoid the implicit resampling that was in some cases
unnecessarily done.
-The optimization for whether an output copy is needed
that was done to improve performance due to the implicit
resampling is no longer needed, which simplifies the
code and makes it less error-prone in the sense that
is no longer neccessary to keep track of whether any
module has changed the signal.
Finally, it should be noted that the polling of the state
for all the submodules was done previously as well, but in
a less obvious and distributed manner.
BUG=webrtc:6181, webrtc:6220, webrtc:5298, webrtc:6296, webrtc:6298, webrtc:6297
Review-Url: https://codereview.webrtc.org/2304123002
Cr-Commit-Position: refs/heads/master@{#14175}
2016-09-10 04:42:27 -07:00
|
|
|
public_submodules_->noise_suppression->is_enabled(),
|
|
|
|
|
capture_nonlocked_.intelligibility_enabled,
|
|
|
|
|
capture_nonlocked_.beamformer_enabled,
|
|
|
|
|
public_submodules_->gain_control->is_enabled(),
|
|
|
|
|
capture_nonlocked_.level_controller_enabled,
|
|
|
|
|
public_submodules_->voice_detection->is_enabled(),
|
|
|
|
|
public_submodules_->level_estimator->is_enabled(),
|
|
|
|
|
capture_.transient_suppressor_enabled);
|
2015-08-14 10:35:55 -07:00
|
|
|
}
|
|
|
|
|
|
2014-12-15 09:41:24 +00:00
|
|
|
|
2015-04-15 11:42:40 +02:00
|
|
|
void AudioProcessingImpl::InitializeTransient() {
|
2015-11-28 12:35:15 -08:00
|
|
|
if (capture_.transient_suppressor_enabled) {
|
|
|
|
|
if (!public_submodules_->transient_suppressor.get()) {
|
|
|
|
|
public_submodules_->transient_suppressor.reset(new TransientSuppressor());
|
2014-12-15 09:41:24 +00:00
|
|
|
}
|
2015-11-28 12:35:15 -08:00
|
|
|
public_submodules_->transient_suppressor->Initialize(
|
2016-09-16 15:02:15 -07:00
|
|
|
capture_nonlocked_.capture_processing_format.sample_rate_hz(),
|
|
|
|
|
capture_nonlocked_.split_rate, num_proc_channels());
|
2014-12-15 09:41:24 +00:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2014-12-19 19:57:34 +00:00
|
|
|
void AudioProcessingImpl::InitializeBeamformer() {
|
2016-01-11 20:32:29 -08:00
|
|
|
if (capture_nonlocked_.beamformer_enabled) {
|
2015-11-28 12:35:15 -08:00
|
|
|
if (!private_submodules_->beamformer) {
|
|
|
|
|
private_submodules_->beamformer.reset(new NonlinearBeamformer(
|
2016-07-01 17:19:09 -07:00
|
|
|
capture_.array_geometry, 1u, capture_.target_direction));
|
2015-01-15 18:07:21 +00:00
|
|
|
}
|
2015-11-28 12:35:15 -08:00
|
|
|
private_submodules_->beamformer->Initialize(kChunkSizeMs,
|
|
|
|
|
capture_nonlocked_.split_rate);
|
2014-12-19 19:57:34 +00:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2015-08-14 10:35:55 -07:00
|
|
|
void AudioProcessingImpl::InitializeIntelligibility() {
|
2016-08-26 07:16:04 -07:00
|
|
|
#if WEBRTC_INTELLIGIBILITY_ENHANCER
|
2016-05-16 15:32:38 -07:00
|
|
|
if (capture_nonlocked_.intelligibility_enabled) {
|
2015-11-28 12:35:15 -08:00
|
|
|
public_submodules_->intelligibility_enhancer.reset(
|
2016-02-22 15:57:38 -08:00
|
|
|
new IntelligibilityEnhancer(capture_nonlocked_.split_rate,
|
2016-03-09 16:24:34 +01:00
|
|
|
render_.render_audio->num_channels(),
|
2016-09-20 14:51:56 -07:00
|
|
|
render_.render_audio->num_bands(),
|
2016-03-09 16:24:34 +01:00
|
|
|
NoiseSuppressionImpl::num_noise_bins()));
|
2015-08-14 10:35:55 -07:00
|
|
|
}
|
2016-08-26 07:16:04 -07:00
|
|
|
#endif
|
2015-08-14 10:35:55 -07:00
|
|
|
}
|
|
|
|
|
|
2016-11-22 07:24:52 -08:00
|
|
|
void AudioProcessingImpl::InitializeLowCutFilter() {
|
|
|
|
|
if (config_.high_pass_filter.enabled) {
|
|
|
|
|
private_submodules_->low_cut_filter.reset(
|
|
|
|
|
new LowCutFilter(num_proc_channels(), proc_sample_rate_hz()));
|
|
|
|
|
} else {
|
|
|
|
|
private_submodules_->low_cut_filter.reset();
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2016-06-29 15:26:12 -07:00
|
|
|
void AudioProcessingImpl::InitializeLevelController() {
|
|
|
|
|
private_submodules_->level_controller->Initialize(proc_sample_rate_hz());
|
|
|
|
|
}
|
|
|
|
|
|
2016-10-28 05:39:16 -07:00
|
|
|
void AudioProcessingImpl::InitializeResidualEchoDetector() {
|
|
|
|
|
private_submodules_->residual_echo_detector->Initialize();
|
|
|
|
|
}
|
|
|
|
|
|
2015-06-29 14:57:29 +02:00
|
|
|
void AudioProcessingImpl::MaybeUpdateHistograms() {
|
2015-07-05 10:46:01 +02:00
|
|
|
static const int kMinDiffDelayMs = 60;
|
2015-06-29 14:57:29 +02:00
|
|
|
|
|
|
|
|
if (echo_cancellation()->is_enabled()) {
|
2015-07-07 11:50:05 +02:00
|
|
|
// Activate delay_jumps_ counters if we know echo_cancellation is runnning.
|
|
|
|
|
// If a stream has echo we know that the echo_cancellation is in process.
|
2015-11-28 12:35:15 -08:00
|
|
|
if (capture_.stream_delay_jumps == -1 &&
|
|
|
|
|
echo_cancellation()->stream_has_echo()) {
|
|
|
|
|
capture_.stream_delay_jumps = 0;
|
2015-07-07 11:50:05 +02:00
|
|
|
}
|
2015-11-28 12:35:15 -08:00
|
|
|
if (capture_.aec_system_delay_jumps == -1 &&
|
2015-07-07 11:50:05 +02:00
|
|
|
echo_cancellation()->stream_has_echo()) {
|
2015-11-28 12:35:15 -08:00
|
|
|
capture_.aec_system_delay_jumps = 0;
|
2015-07-07 11:50:05 +02:00
|
|
|
}
|
|
|
|
|
|
2015-06-29 14:57:29 +02:00
|
|
|
// Detect a jump in platform reported system delay and log the difference.
|
2015-11-28 12:35:15 -08:00
|
|
|
const int diff_stream_delay_ms =
|
|
|
|
|
capture_nonlocked_.stream_delay_ms - capture_.last_stream_delay_ms;
|
|
|
|
|
if (diff_stream_delay_ms > kMinDiffDelayMs &&
|
|
|
|
|
capture_.last_stream_delay_ms != 0) {
|
2016-03-07 01:52:59 -08:00
|
|
|
RTC_HISTOGRAM_COUNTS("WebRTC.Audio.PlatformReportedStreamDelayJump",
|
|
|
|
|
diff_stream_delay_ms, kMinDiffDelayMs, 1000, 100);
|
2015-11-28 12:35:15 -08:00
|
|
|
if (capture_.stream_delay_jumps == -1) {
|
|
|
|
|
capture_.stream_delay_jumps = 0; // Activate counter if needed.
|
2015-07-07 11:50:05 +02:00
|
|
|
}
|
2015-11-28 12:35:15 -08:00
|
|
|
capture_.stream_delay_jumps++;
|
2015-06-29 14:57:29 +02:00
|
|
|
}
|
2015-11-28 12:35:15 -08:00
|
|
|
capture_.last_stream_delay_ms = capture_nonlocked_.stream_delay_ms;
|
2015-06-29 14:57:29 +02:00
|
|
|
|
|
|
|
|
// Detect a jump in AEC system delay and log the difference.
|
2016-03-04 11:50:54 -08:00
|
|
|
const int samples_per_ms =
|
2015-11-28 12:35:15 -08:00
|
|
|
rtc::CheckedDivExact(capture_nonlocked_.split_rate, 1000);
|
2016-03-04 11:50:54 -08:00
|
|
|
RTC_DCHECK_LT(0, samples_per_ms);
|
2015-06-29 14:57:29 +02:00
|
|
|
const int aec_system_delay_ms =
|
2016-03-04 11:50:54 -08:00
|
|
|
public_submodules_->echo_cancellation->GetSystemDelayInSamples() /
|
|
|
|
|
samples_per_ms;
|
2015-07-23 11:41:39 -07:00
|
|
|
const int diff_aec_system_delay_ms =
|
2015-11-28 12:35:15 -08:00
|
|
|
aec_system_delay_ms - capture_.last_aec_system_delay_ms;
|
2015-06-29 14:57:29 +02:00
|
|
|
if (diff_aec_system_delay_ms > kMinDiffDelayMs &&
|
2015-11-28 12:35:15 -08:00
|
|
|
capture_.last_aec_system_delay_ms != 0) {
|
2016-03-07 01:52:59 -08:00
|
|
|
RTC_HISTOGRAM_COUNTS("WebRTC.Audio.AecSystemDelayJump",
|
|
|
|
|
diff_aec_system_delay_ms, kMinDiffDelayMs, 1000,
|
|
|
|
|
100);
|
2015-11-28 12:35:15 -08:00
|
|
|
if (capture_.aec_system_delay_jumps == -1) {
|
|
|
|
|
capture_.aec_system_delay_jumps = 0; // Activate counter if needed.
|
2015-07-07 11:50:05 +02:00
|
|
|
}
|
2015-11-28 12:35:15 -08:00
|
|
|
capture_.aec_system_delay_jumps++;
|
2015-06-29 14:57:29 +02:00
|
|
|
}
|
2015-11-28 12:35:15 -08:00
|
|
|
capture_.last_aec_system_delay_ms = aec_system_delay_ms;
|
2015-06-29 14:57:29 +02:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2015-07-07 11:50:05 +02:00
|
|
|
void AudioProcessingImpl::UpdateHistogramsOnCallEnd() {
|
2015-11-28 12:35:15 -08:00
|
|
|
// Run in a single-threaded manner.
|
|
|
|
|
rtc::CritScope cs_render(&crit_render_);
|
|
|
|
|
rtc::CritScope cs_capture(&crit_capture_);
|
|
|
|
|
|
|
|
|
|
if (capture_.stream_delay_jumps > -1) {
|
2016-03-07 01:52:59 -08:00
|
|
|
RTC_HISTOGRAM_ENUMERATION(
|
2015-07-07 11:50:05 +02:00
|
|
|
"WebRTC.Audio.NumOfPlatformReportedStreamDelayJumps",
|
2015-11-28 12:35:15 -08:00
|
|
|
capture_.stream_delay_jumps, 51);
|
2015-07-07 11:50:05 +02:00
|
|
|
}
|
2015-11-28 12:35:15 -08:00
|
|
|
capture_.stream_delay_jumps = -1;
|
|
|
|
|
capture_.last_stream_delay_ms = 0;
|
2015-07-07 11:50:05 +02:00
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
if (capture_.aec_system_delay_jumps > -1) {
|
2016-03-07 01:52:59 -08:00
|
|
|
RTC_HISTOGRAM_ENUMERATION("WebRTC.Audio.NumOfAecSystemDelayJumps",
|
|
|
|
|
capture_.aec_system_delay_jumps, 51);
|
2015-07-07 11:50:05 +02:00
|
|
|
}
|
2015-11-28 12:35:15 -08:00
|
|
|
capture_.aec_system_delay_jumps = -1;
|
|
|
|
|
capture_.last_aec_system_delay_ms = 0;
|
2015-07-07 11:50:05 +02:00
|
|
|
}
|
|
|
|
|
|
2011-12-03 00:03:31 +00:00
|
|
|
#ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP
|
2015-11-28 12:35:15 -08:00
|
|
|
int AudioProcessingImpl::WriteMessageToDebugFile(
|
|
|
|
|
FileWrapper* debug_file,
|
2016-01-15 03:06:36 -08:00
|
|
|
int64_t* filesize_limit_bytes,
|
2015-11-28 12:35:15 -08:00
|
|
|
rtc::CriticalSection* crit_debug,
|
|
|
|
|
ApmDebugDumpThreadState* debug_state) {
|
|
|
|
|
int32_t size = debug_state->event_msg->ByteSize();
|
2011-08-03 21:08:51 +00:00
|
|
|
if (size <= 0) {
|
|
|
|
|
return kUnspecifiedError;
|
|
|
|
|
}
|
2013-10-22 10:27:23 +00:00
|
|
|
#if defined(WEBRTC_ARCH_BIG_ENDIAN)
|
2015-07-23 11:41:39 -07:00
|
|
|
// TODO(ajm): Use little-endian "on the wire". For the moment, we can be
|
|
|
|
|
// pretty safe in assuming little-endian.
|
2011-08-03 21:08:51 +00:00
|
|
|
#endif
|
|
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
if (!debug_state->event_msg->SerializeToString(&debug_state->event_str)) {
|
2011-08-03 21:08:51 +00:00
|
|
|
return kUnspecifiedError;
|
|
|
|
|
}
|
|
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
{
|
|
|
|
|
// Ensure atomic writes of the message.
|
2016-01-15 03:06:36 -08:00
|
|
|
rtc::CritScope cs_debug(crit_debug);
|
|
|
|
|
|
FileWrapper[Impl] modifications and actually remove the "Impl" class.
This is a somewhat involved refactoring of this class. Here's an overview of the changes:
* FileWrapper can now be used as a regular class and instances allocated on the stack.
* The type now has support for move semantics and copy isn't allowed.
* New public ctor with FILE* that can be used instead of OpenFromFileHandle.
* New static Open() method. The intent of this is to allow opening a file and getting back a FileWrapper instance. Using this method instead of Create(), will allow us in the future to make the FILE* member pointer, to be const and simplify threading (get rid of the lock).
* Rename the Open() method to is_open() and make it inline.
* The FileWrapper interface is no longer a pure virtual interface. There's only one implementation so there's no need to go through a vtable for everything.
* Functionality offered by the class, is now reduced. No support for looping (not clear if that was actually useful to users of that flag), no need to implement the 'read_only_' functionality in the class, since file APIs implement that already, no support for *not* managing the file handle (this wasn't used). OpenFromFileHandle always "manages" the file.
* Delete the unused WriteText() method and don't support opening files in text mode. Text mode is only different on Windows and on Windows it translates \n to \r\n, which means that files such as log files, could have a slightly different format on Windows than other platforms. Besides, tools on Windows can handle UNIX line endings.
* Remove FileName(), change Trace code to manage its own path.
* Rename id_ member variable to file_.
* Removed the open_ member variable since the same functionality can be gotten from just checking the file pointer.
* Don't call CloseFile inside of Write. Write shouldn't be changing the state of the class beyond just attempting to write.
* Remove concept of looping from FileWrapper and never close inside of Read()
* Changed stream base classes to inherit from a common base class instead of both defining the Rewind method. Ultimately, Id' like to remove these interfaces and just have FileWrapper.
* Remove read_only param from OpenFromFileHandle
* Renamed size_in_bytes_ to position_, since it gets set to 0 when Rewind() is called (and the size actually does not change).
* Switch out rw lock for CriticalSection. The r/w lock was only used for reading when checking the open_ flag.
BUG=
Review-Url: https://codereview.webrtc.org/2054373002
Cr-Commit-Position: refs/heads/master@{#13155}
2016-06-15 10:30:14 -07:00
|
|
|
RTC_DCHECK(debug_file->is_open());
|
2016-01-15 03:06:36 -08:00
|
|
|
// Update the byte counter.
|
|
|
|
|
if (*filesize_limit_bytes >= 0) {
|
|
|
|
|
*filesize_limit_bytes -=
|
|
|
|
|
(sizeof(int32_t) + debug_state->event_str.length());
|
|
|
|
|
if (*filesize_limit_bytes < 0) {
|
|
|
|
|
// Not enough bytes are left to write this message, so stop logging.
|
|
|
|
|
debug_file->CloseFile();
|
|
|
|
|
return kNoError;
|
|
|
|
|
}
|
|
|
|
|
}
|
2015-11-28 12:35:15 -08:00
|
|
|
// Write message preceded by its size.
|
|
|
|
|
if (!debug_file->Write(&size, sizeof(int32_t))) {
|
|
|
|
|
return kFileError;
|
|
|
|
|
}
|
|
|
|
|
if (!debug_file->Write(debug_state->event_str.data(),
|
|
|
|
|
debug_state->event_str.length())) {
|
|
|
|
|
return kFileError;
|
|
|
|
|
}
|
2011-08-03 21:08:51 +00:00
|
|
|
}
|
|
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
debug_state->event_msg->Clear();
|
2011-08-03 21:08:51 +00:00
|
|
|
|
2014-03-04 20:58:13 +00:00
|
|
|
return kNoError;
|
2011-08-03 21:08:51 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
int AudioProcessingImpl::WriteInitMessage() {
|
2015-11-28 12:35:15 -08:00
|
|
|
debug_dump_.capture.event_msg->set_type(audioproc::Event::INIT);
|
|
|
|
|
audioproc::Init* msg = debug_dump_.capture.event_msg->mutable_init();
|
|
|
|
|
msg->set_sample_rate(formats_.api_format.input_stream().sample_rate_hz());
|
|
|
|
|
|
Convert channel counts to size_t.
IIRC, this was originally requested by ajm during review of the other size_t conversions I did over the past year, and I agreed it made sense, but wanted to do it separately since those changes were already gargantuan.
BUG=chromium:81439
TEST=none
R=henrik.lundin@webrtc.org, henrika@webrtc.org, kjellander@webrtc.org, minyue@webrtc.org, perkj@webrtc.org, solenberg@webrtc.org, stefan@webrtc.org, tina.legrand@webrtc.org
Review URL: https://codereview.webrtc.org/1316523002 .
Cr-Commit-Position: refs/heads/master@{#11229}
2016-01-12 16:26:35 -08:00
|
|
|
msg->set_num_input_channels(static_cast<google::protobuf::int32>(
|
|
|
|
|
formats_.api_format.input_stream().num_channels()));
|
|
|
|
|
msg->set_num_output_channels(static_cast<google::protobuf::int32>(
|
|
|
|
|
formats_.api_format.output_stream().num_channels()));
|
|
|
|
|
msg->set_num_reverse_channels(static_cast<google::protobuf::int32>(
|
|
|
|
|
formats_.api_format.reverse_input_stream().num_channels()));
|
2015-08-14 10:35:55 -07:00
|
|
|
msg->set_reverse_sample_rate(
|
2015-11-28 12:35:15 -08:00
|
|
|
formats_.api_format.reverse_input_stream().sample_rate_hz());
|
2015-11-17 02:16:45 -08:00
|
|
|
msg->set_output_sample_rate(
|
2015-11-28 12:35:15 -08:00
|
|
|
formats_.api_format.output_stream().sample_rate_hz());
|
2016-04-11 07:05:53 -07:00
|
|
|
msg->set_reverse_output_sample_rate(
|
|
|
|
|
formats_.api_format.reverse_output_stream().sample_rate_hz());
|
|
|
|
|
msg->set_num_reverse_output_channels(
|
|
|
|
|
formats_.api_format.reverse_output_stream().num_channels());
|
2011-08-03 21:08:51 +00:00
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
RETURN_ON_ERR(WriteMessageToDebugFile(debug_dump_.debug_file.get(),
|
2016-01-15 03:06:36 -08:00
|
|
|
&debug_dump_.num_bytes_left_for_log_,
|
2015-11-28 12:35:15 -08:00
|
|
|
&crit_debug_, &debug_dump_.capture));
|
2015-10-03 00:39:14 +02:00
|
|
|
return kNoError;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
int AudioProcessingImpl::WriteConfigMessage(bool forced) {
|
|
|
|
|
audioproc::Config config;
|
|
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
config.set_aec_enabled(public_submodules_->echo_cancellation->is_enabled());
|
2015-10-03 00:39:14 +02:00
|
|
|
config.set_aec_delay_agnostic_enabled(
|
2015-11-28 12:35:15 -08:00
|
|
|
public_submodules_->echo_cancellation->is_delay_agnostic_enabled());
|
2015-10-03 00:39:14 +02:00
|
|
|
config.set_aec_drift_compensation_enabled(
|
2015-11-28 12:35:15 -08:00
|
|
|
public_submodules_->echo_cancellation->is_drift_compensation_enabled());
|
2015-10-03 00:39:14 +02:00
|
|
|
config.set_aec_extended_filter_enabled(
|
2015-11-28 12:35:15 -08:00
|
|
|
public_submodules_->echo_cancellation->is_extended_filter_enabled());
|
|
|
|
|
config.set_aec_suppression_level(static_cast<int>(
|
|
|
|
|
public_submodules_->echo_cancellation->suppression_level()));
|
2015-10-03 00:39:14 +02:00
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
config.set_aecm_enabled(
|
|
|
|
|
public_submodules_->echo_control_mobile->is_enabled());
|
2015-10-03 00:39:14 +02:00
|
|
|
config.set_aecm_comfort_noise_enabled(
|
2015-11-28 12:35:15 -08:00
|
|
|
public_submodules_->echo_control_mobile->is_comfort_noise_enabled());
|
|
|
|
|
config.set_aecm_routing_mode(static_cast<int>(
|
|
|
|
|
public_submodules_->echo_control_mobile->routing_mode()));
|
2015-10-03 00:39:14 +02:00
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
config.set_agc_enabled(public_submodules_->gain_control->is_enabled());
|
|
|
|
|
config.set_agc_mode(
|
|
|
|
|
static_cast<int>(public_submodules_->gain_control->mode()));
|
|
|
|
|
config.set_agc_limiter_enabled(
|
|
|
|
|
public_submodules_->gain_control->is_limiter_enabled());
|
2016-02-13 16:40:47 -08:00
|
|
|
config.set_noise_robust_agc_enabled(constants_.use_experimental_agc);
|
2015-10-03 00:39:14 +02:00
|
|
|
|
2016-11-22 07:24:52 -08:00
|
|
|
config.set_hpf_enabled(config_.high_pass_filter.enabled);
|
2015-10-03 00:39:14 +02:00
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
config.set_ns_enabled(public_submodules_->noise_suppression->is_enabled());
|
|
|
|
|
config.set_ns_level(
|
|
|
|
|
static_cast<int>(public_submodules_->noise_suppression->level()));
|
2015-10-03 00:39:14 +02:00
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
config.set_transient_suppression_enabled(
|
|
|
|
|
capture_.transient_suppressor_enabled);
|
2016-05-16 15:32:38 -07:00
|
|
|
config.set_intelligibility_enhancer_enabled(
|
|
|
|
|
capture_nonlocked_.intelligibility_enabled);
|
2015-10-03 00:39:14 +02:00
|
|
|
|
2016-04-15 01:19:44 -07:00
|
|
|
std::string experiments_description =
|
|
|
|
|
public_submodules_->echo_cancellation->GetExperimentsDescription();
|
|
|
|
|
// TODO(peah): Add semicolon-separated concatenations of experiment
|
|
|
|
|
// descriptions for other submodules.
|
2016-06-29 15:26:12 -07:00
|
|
|
if (capture_nonlocked_.level_controller_enabled) {
|
|
|
|
|
experiments_description += "LevelController;";
|
|
|
|
|
}
|
2016-04-15 01:19:44 -07:00
|
|
|
config.set_experiments_description(experiments_description);
|
|
|
|
|
|
2015-10-03 00:39:14 +02:00
|
|
|
std::string serialized_config = config.SerializeAsString();
|
2015-11-28 12:35:15 -08:00
|
|
|
if (!forced &&
|
|
|
|
|
debug_dump_.capture.last_serialized_config == serialized_config) {
|
2015-10-03 00:39:14 +02:00
|
|
|
return kNoError;
|
2011-08-03 21:08:51 +00:00
|
|
|
}
|
|
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
debug_dump_.capture.last_serialized_config = serialized_config;
|
2015-10-03 00:39:14 +02:00
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
debug_dump_.capture.event_msg->set_type(audioproc::Event::CONFIG);
|
|
|
|
|
debug_dump_.capture.event_msg->mutable_config()->CopyFrom(config);
|
2015-10-03 00:39:14 +02:00
|
|
|
|
2015-11-28 12:35:15 -08:00
|
|
|
RETURN_ON_ERR(WriteMessageToDebugFile(debug_dump_.debug_file.get(),
|
2016-01-15 03:06:36 -08:00
|
|
|
&debug_dump_.num_bytes_left_for_log_,
|
2015-11-28 12:35:15 -08:00
|
|
|
&crit_debug_, &debug_dump_.capture));
|
2011-08-03 21:08:51 +00:00
|
|
|
return kNoError;
|
|
|
|
|
}
|
2011-12-03 00:03:31 +00:00
|
|
|
#endif // WEBRTC_AUDIOPROC_DEBUG_DUMP
|
2014-04-22 21:00:04 +00:00
|
|
|
|
2016-08-29 14:46:07 -07:00
|
|
|
AudioProcessingImpl::ApmCaptureState::ApmCaptureState(
|
|
|
|
|
bool transient_suppressor_enabled,
|
|
|
|
|
const std::vector<Point>& array_geometry,
|
|
|
|
|
SphericalPointf target_direction)
|
|
|
|
|
: aec_system_delay_jumps(-1),
|
|
|
|
|
delay_offset_ms(0),
|
|
|
|
|
was_stream_delay_set(false),
|
|
|
|
|
last_stream_delay_ms(0),
|
|
|
|
|
last_aec_system_delay_ms(0),
|
|
|
|
|
stream_delay_jumps(-1),
|
|
|
|
|
output_will_be_muted(false),
|
|
|
|
|
key_pressed(false),
|
|
|
|
|
transient_suppressor_enabled(transient_suppressor_enabled),
|
|
|
|
|
array_geometry(array_geometry),
|
|
|
|
|
target_direction(target_direction),
|
2016-09-16 15:02:15 -07:00
|
|
|
capture_processing_format(kSampleRate16kHz),
|
2016-08-29 14:46:07 -07:00
|
|
|
split_rate(kSampleRate16kHz) {}
|
|
|
|
|
|
|
|
|
|
AudioProcessingImpl::ApmCaptureState::~ApmCaptureState() = default;
|
|
|
|
|
|
|
|
|
|
AudioProcessingImpl::ApmRenderState::ApmRenderState() = default;
|
|
|
|
|
|
|
|
|
|
AudioProcessingImpl::ApmRenderState::~ApmRenderState() = default;
|
|
|
|
|
|
2011-07-07 08:21:25 +00:00
|
|
|
} // namespace webrtc
|