New module for the adaptive level controlling functionality in the audio processing module

webrtc/modules/audio_processing/audio_processing_impl.cc

Index: webrtc/modules/audio_processing/audio_processing_impl.cc
diff --git a/webrtc/modules/audio_processing/audio_processing_impl.cc b/webrtc/modules/audio_processing/audio_processing_impl.cc
index 819a18b62d958f38e7b4a5f6b5f92687c8cf59bb..66edad374211b9aaaf533a2f37f76467d1de9ed0 100644
--- a/webrtc/modules/audio_processing/audio_processing_impl.cc
+++ b/webrtc/modules/audio_processing/audio_processing_impl.cc
@@ -31,6 +31,7 @@
 #include "webrtc/modules/audio_processing/gain_control_impl.h"
 #include "webrtc/modules/audio_processing/high_pass_filter_impl.h"
 #include "webrtc/modules/audio_processing/intelligibility/intelligibility_enhancer.h"
+#include "webrtc/modules/audio_processing/level_controller/level_controller.h"
 #include "webrtc/modules/audio_processing/level_estimator_impl.h"
 #include "webrtc/modules/audio_processing/noise_suppression_impl.h"
 #include "webrtc/modules/audio_processing/transient/transient_suppressor.h"
@@ -132,6 +133,7 @@ struct AudioProcessingImpl::ApmPrivateSubmodules {
   // Accessed internally from capture or during initialization
   std::unique_ptr<Beamformer<float>> beamformer;
   std::unique_ptr<AgcManagerDirect> agc_manager;
+  std::unique_ptr<LevelController> level_controller;
 };
 
 AudioProcessing* AudioProcessing::Create() {
@@ -175,8 +177,8 @@ AudioProcessingImpl::AudioProcessingImpl(const Config& config,
                config.Get<Beamforming>().array_geometry,
                config.Get<Beamforming>().target_direction),
       capture_nonlocked_(config.Get<Beamforming>().enabled,
-                         config.Get<Intelligibility>().enabled)
-{
+                         config.Get<Intelligibility>().enabled,
+                         config.Get<LevelControl>().enabled) {
   {
     rtc::CritScope cs_render(&crit_render_);
     rtc::CritScope cs_capture(&crit_capture_);
@@ -198,6 +200,8 @@ AudioProcessingImpl::AudioProcessingImpl(const Config& config,
     public_submodules_->gain_control_for_experimental_agc.reset(
         new GainControlForExperimentalAgc(
             public_submodules_->gain_control.get(), &crit_capture_));
+
+    private_submodules_->level_controller.reset(new LevelController());
   }
 
   SetExtraOptions(config);
@@ -322,6 +326,7 @@ int AudioProcessingImpl::InitializeLocked() {
   InitializeNoiseSuppression();
   InitializeLevelEstimator();
   InitializeVoiceDetection();
+  InitializeLevelController();
 
 #ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP
   if (debug_dump_.debug_file->is_open()) {
@@ -408,6 +413,13 @@ void AudioProcessingImpl::SetExtraOptions(const Config& config) {
     InitializeTransient();
   }
 
+  if (capture_nonlocked_.level_controller_enabled !=
+      config.Get<LevelControl>().enabled) {
+    capture_nonlocked_.level_controller_enabled =
+        config.Get<LevelControl>().enabled;
+    InitializeLevelController();
+  }
+
   if(capture_nonlocked_.intelligibility_enabled !=
      config.Get<Intelligibility>().enabled) {
     capture_nonlocked_.intelligibility_enabled =
@@ -759,6 +771,10 @@ int AudioProcessingImpl::ProcessStreamLocked() {
         capture_.key_pressed);
   }
 
+  if (capture_nonlocked_.level_controller_enabled) {
+    private_submodules_->level_controller->Process(ca);

[aluebs-webrtc, 2016/06/28 22:45:02]

I am not sure how this component works (it is not in the CL description or
documented in the header file), but if it applies an additional gain to the
signal, we probably want to compensate for that in the gain_db parameter when
calling intelligibility_enhancer->SetCaptureNoiseEstimate() in line 728.

[peah-webrtc, 2016/06/29 06:22:09]

The component indeed applies an additional gain. 

I'll add an API call for obtaining the latest applied gain, but I'll let you use
that gain in the intelligibility enhancer in another CL since
1) I don't really understand how that gain is to be used in the intelligibility
enhancer.
2) It is nice to make this CL only concern the level control and not also affect
the intelligibility enhancer.
3) The intelligibility enhancer is currently not active by default so the
addition of the level control will not cause any downstream breakages.

Lets discuss it further offline if you want.

[aluebs-webrtc, 2016/06/29 16:40:28]

2) This CL concerns the IE, because it applies a gain without compensating for
it.
3) It can be manually enabled and some of us are using it already.
But I can do a separate CL if you want.

[peah-webrtc, 2016/06/29 20:39:59]

I definitely see that it affects the IE, but I think it is better if you do that
change as you know how that is designed. Furthermore, I'd expect that that
change needs to be verified carefully, as the level controller may change the
power of the signal by as much as 20 dB.

An alternative is to ensure that the IE and the level control are never active
at the same time.

+  }
+
   // The level estimator operates on the recombined data.
   public_submodules_->level_estimator->ProcessStream(ca);
 
@@ -1118,7 +1134,8 @@ bool AudioProcessingImpl::output_copy_needed() const {
   // Check if we've upmixed or downmixed the audio.
   return ((formats_.api_format.output_stream().num_channels() !=
            formats_.api_format.input_stream().num_channels()) ||
-          is_fwd_processed() || capture_.transient_suppressor_enabled);
+          is_fwd_processed() || capture_.transient_suppressor_enabled ||
+          capture_nonlocked_.level_controller_enabled);
 }
 
 bool AudioProcessingImpl::fwd_synthesis_needed() const {
@@ -1247,6 +1264,10 @@ void AudioProcessingImpl::InitializeLevelEstimator() {
   public_submodules_->level_estimator->Initialize();
 }
 
+void AudioProcessingImpl::InitializeLevelController() {
+  private_submodules_->level_controller->Initialize(proc_sample_rate_hz());
+}
+
 void AudioProcessingImpl::InitializeVoiceDetection() {
   public_submodules_->voice_detection->Initialize(proc_split_sample_rate_hz());
 }