Delete voice_detection() pointer to submodule
The new configuration path is via AudioProcessing::ApplyConfig and
AudioProcessing::GetStatistics.
ApmTest.Process passes with unchanged reference files if
audio_processing_impl would initialize the VAD with
VoiceDetection::kLowLikelihood instead of kVeryLowLikelihood.
This was verified by testing this CL with that modification.
Bug: webrtc:9878
Change-Id: I4d08df37a07e5c72feeec02a07d6b9435f917d72
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/155445
Commit-Queue: Sam Zackrisson <saza@webrtc.org>
Reviewed-by: Ivo Creusen <ivoc@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29395}
diff --git a/modules/audio_processing/audio_processing_impl.cc b/modules/audio_processing/audio_processing_impl.cc
index b1187fa..c661848 100644
--- a/modules/audio_processing/audio_processing_impl.cc
+++ b/modules/audio_processing/audio_processing_impl.cc
@@ -40,7 +40,7 @@
#include "modules/audio_processing/noise_suppression_proxy.h"
#include "modules/audio_processing/residual_echo_detector.h"
#include "modules/audio_processing/transient/transient_suppressor.h"
-#include "modules/audio_processing/voice_detection_impl.h"
+#include "modules/audio_processing/voice_detection.h"
#include "rtc_base/atomic_ops.h"
#include "rtc_base/checks.h"
#include "rtc_base/constructor_magic.h"
@@ -165,8 +165,7 @@
bool gain_controller2_enabled,
bool pre_amplifier_enabled,
bool echo_controller_enabled,
- bool voice_activity_detector_enabled,
- bool private_voice_detector_enabled,
+ bool voice_detector_enabled,
bool level_estimator_enabled,
bool transient_suppressor_enabled) {
bool changed = false;
@@ -183,10 +182,7 @@
changed |= (pre_amplifier_enabled_ != pre_amplifier_enabled);
changed |= (echo_controller_enabled != echo_controller_enabled_);
changed |= (level_estimator_enabled != level_estimator_enabled_);
- changed |=
- (voice_activity_detector_enabled != voice_activity_detector_enabled_);
- changed |=
- (private_voice_detector_enabled != private_voice_detector_enabled_);
+ changed |= (voice_detector_enabled != voice_detector_enabled_);
changed |= (transient_suppressor_enabled != transient_suppressor_enabled_);
if (changed) {
high_pass_filter_enabled_ = high_pass_filter_enabled;
@@ -199,8 +195,7 @@
pre_amplifier_enabled_ = pre_amplifier_enabled;
echo_controller_enabled_ = echo_controller_enabled;
level_estimator_enabled_ = level_estimator_enabled;
- voice_activity_detector_enabled_ = voice_activity_detector_enabled;
- private_voice_detector_enabled_ = private_voice_detector_enabled;
+ voice_detector_enabled_ = voice_detector_enabled;
transient_suppressor_enabled_ = transient_suppressor_enabled;
}
@@ -211,8 +206,7 @@
bool AudioProcessingImpl::ApmSubmoduleStates::CaptureMultiBandSubModulesActive()
const {
- return CaptureMultiBandProcessingActive() ||
- voice_activity_detector_enabled_ || private_voice_detector_enabled_;
+ return CaptureMultiBandProcessingActive() || voice_detector_enabled_;
}
bool AudioProcessingImpl::ApmSubmoduleStates::CaptureMultiBandProcessingActive()
@@ -263,7 +257,6 @@
std::unique_ptr<LevelEstimatorImpl> level_estimator;
std::unique_ptr<NoiseSuppressionImpl> noise_suppression;
std::unique_ptr<NoiseSuppressionProxy> noise_suppression_proxy;
- std::unique_ptr<VoiceDetectionImpl> voice_detection;
std::unique_ptr<GainControlImpl> gain_control;
std::unique_ptr<GainControlForExperimentalAgc>
gain_control_for_experimental_agc;
@@ -295,7 +288,7 @@
std::unique_ptr<GainApplier> pre_amplifier;
std::unique_ptr<CustomAudioAnalyzer> capture_analyzer;
std::unique_ptr<LevelEstimatorImpl> output_level_estimator;
- std::unique_ptr<VoiceDetectionImpl> voice_detector;
+ std::unique_ptr<VoiceDetection> voice_detector;
};
AudioProcessingBuilder::AudioProcessingBuilder() = default;
@@ -415,8 +408,6 @@
new NoiseSuppressionImpl(&crit_capture_));
public_submodules_->noise_suppression_proxy.reset(new NoiseSuppressionProxy(
this, public_submodules_->noise_suppression.get()));
- public_submodules_->voice_detection.reset(
- new VoiceDetectionImpl(&crit_capture_));
public_submodules_->gain_control_for_experimental_agc.reset(
new GainControlForExperimentalAgc(
public_submodules_->gain_control.get()));
@@ -556,11 +547,7 @@
InitializeHighPassFilter();
public_submodules_->noise_suppression->Initialize(num_proc_channels(),
proc_sample_rate_hz());
- public_submodules_->voice_detection->Initialize(proc_split_sample_rate_hz());
- if (private_submodules_->voice_detector) {
- private_submodules_->voice_detector->Initialize(
- proc_split_sample_rate_hz());
- }
+ InitializeVoiceDetector();
public_submodules_->level_estimator->Initialize();
InitializeResidualEchoDetector();
InitializeEchoController();
@@ -702,6 +689,9 @@
config_.gain_controller1.analog_level_maximum !=
config.gain_controller1.analog_level_maximum;
+ const bool voice_detection_config_changed =
+ config_.voice_detection.enabled != config.voice_detection.enabled;
+
config_ = config;
if (aec_config_changed) {
@@ -745,14 +735,8 @@
private_submodules_->output_level_estimator->Enable(true);
}
- if (config_.voice_detection.enabled && !private_submodules_->voice_detector) {
- private_submodules_->voice_detector.reset(
- new VoiceDetectionImpl(&crit_capture_));
- private_submodules_->voice_detector->Enable(true);
- private_submodules_->voice_detector->set_likelihood(
- VoiceDetection::kVeryLowLikelihood);
- private_submodules_->voice_detector->Initialize(
- proc_split_sample_rate_hz());
+ if (voice_detection_config_changed) {
+ InitializeVoiceDetector();
}
// Reinitialization must happen after all submodule configuration to avoid
@@ -1276,14 +1260,17 @@
RecordUnprocessedCaptureStream(*frame);
}
- capture_.vad_activity = frame->vad_activity_;
capture_.capture_audio->CopyFrom(frame);
RETURN_ON_ERR(ProcessCaptureStreamLocked());
if (submodule_states_.CaptureMultiBandProcessingActive() ||
submodule_states_.CaptureFullBandProcessingActive()) {
capture_.capture_audio->CopyTo(frame);
}
- frame->vad_activity_ = capture_.vad_activity;
+ if (capture_.stats.voice_detected) {
+ frame->vad_activity_ = *capture_.stats.voice_detected
+ ? AudioFrame::kVadActive
+ : AudioFrame::kVadPassive;
+ }
if (aec_dump_) {
RecordProcessedCaptureStream(*frame);
@@ -1432,19 +1419,10 @@
public_submodules_->noise_suppression->ProcessCaptureAudio(capture_buffer);
}
- if (public_submodules_->voice_detection->is_enabled() &&
- !public_submodules_->voice_detection->using_external_vad()) {
- bool voice_active =
- public_submodules_->voice_detection->ProcessCaptureAudio(
- capture_buffer);
- capture_.vad_activity =
- voice_active ? AudioFrame::kVadActive : AudioFrame::kVadPassive;
- }
-
if (config_.voice_detection.enabled) {
- private_submodules_->voice_detector->ProcessCaptureAudio(capture_buffer);
capture_.stats.voice_detected =
- private_submodules_->voice_detector->stream_has_voice();
+ private_submodules_->voice_detector->ProcessCaptureAudio(
+ capture_buffer);
} else {
capture_.stats.voice_detected = absl::nullopt;
}
@@ -1817,10 +1795,6 @@
return public_submodules_->noise_suppression_proxy.get();
}
-VoiceDetection* AudioProcessingImpl::voice_detection() const {
- return public_submodules_->voice_detection.get();
-}
-
void AudioProcessingImpl::MutateConfig(
rtc::FunctionView<void(AudioProcessing::Config*)> mutator) {
rtc::CritScope cs_render(&crit_render_);
@@ -1845,7 +1819,6 @@
public_submodules_->gain_control->is_enabled(),
config_.gain_controller2.enabled, config_.pre_amplifier.enabled,
capture_nonlocked_.echo_controller_enabled,
- public_submodules_->voice_detection->is_enabled(),
config_.voice_detection.enabled,
public_submodules_->level_estimator->is_enabled(),
capture_.transient_suppressor_enabled);
@@ -1871,6 +1844,14 @@
}
}
+void AudioProcessingImpl::InitializeVoiceDetector() {
+ if (config_.voice_detection.enabled) {
+ private_submodules_->voice_detector = std::make_unique<VoiceDetection>(
+ proc_split_sample_rate_hz(), VoiceDetection::kVeryLowLikelihood);
+ } else {
+ private_submodules_->voice_detector.reset();
+ }
+}
void AudioProcessingImpl::InitializeEchoController() {
bool use_echo_controller =
echo_control_factory_ ||