Integrate Intelligibility with APM

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 bbfb771182cc7196f0e8e0ca9fb63860617d6b8f..dbfa336a6a9a5f24735e8d7298a6cc73d43730e8 100644
--- a/webrtc/modules/audio_processing/audio_processing_impl.cc
+++ b/webrtc/modules/audio_processing/audio_processing_impl.cc
@@ -29,6 +29,7 @@ extern "C" {
 #include "webrtc/modules/audio_processing/echo_control_mobile_impl.h"
 #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_estimator_impl.h"
 #include "webrtc/modules/audio_processing/noise_suppression_impl.h"
 #include "webrtc/modules/audio_processing/processing_component.h"
@@ -210,7 +211,8 @@ AudioProcessingImpl::AudioProcessingImpl(const Config& config,
 #endif
       beamformer_enabled_(config.Get<Beamforming>().enabled),
       beamformer_(beamformer),
-      array_geometry_(config.Get<Beamforming>().array_geometry) {
+      array_geometry_(config.Get<Beamforming>().array_geometry),
+      intelligibility_enabled_(config.Get<Intelligibility>().enabled) {
   echo_cancellation_ = new EchoCancellationImpl(this, crit_);
   component_list_.push_back(echo_cancellation_);
 
@@ -329,6 +331,8 @@ int AudioProcessingImpl::InitializeLocked() {
 
   InitializeBeamformer();
 
+  InitializeIntelligibility();
+
 #ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP
   if (debug_file_->Open()) {
     int err = WriteInitMessage();
@@ -623,6 +627,7 @@ int AudioProcessingImpl::ProcessStreamLocked() {
   MaybeUpdateHistograms();
 
   AudioBuffer* ca = capture_audio_.get();  // For brevity.
+
   if (use_new_agc_ && gain_control_->is_enabled()) {
     agc_manager_->AnalyzePreProcess(ca->channels()[0], ca->num_channels(),
                                     fwd_proc_format_.num_frames());
@@ -633,6 +638,11 @@ int AudioProcessingImpl::ProcessStreamLocked() {
     ca->SplitIntoFrequencyBands();
   }
 
+  if (intelligibility_enabled_) {
+    intelligibility_enhancer_->AnalyzeCaptureAudio(
+        ca->split_channels_f(kBand0To8kHz), split_rate_, ca->num_channels());
+  }
+
   if (beamformer_enabled_) {
     beamformer_->ProcessChunk(*ca->split_data_f(), ca->split_data_f());
     ca->set_num_channels(1);
@@ -681,12 +691,25 @@ int AudioProcessingImpl::ProcessStreamLocked() {
   return kNoError;
 }
 
+int AudioProcessingImpl::ProcessReverseStream(float* const* data,
+                                              int samples_per_channel,
+                                              int rev_sample_rate_hz,
+                                              ChannelLayout layout) {
+  RETURN_ON_ERR(AnalyzeReverseStream(data, samples_per_channel,
+                                     rev_sample_rate_hz, layout));
+  if (is_rev_processed()) {
+    render_audio_->CopyTo(api_format_.reverse_stream(), data);
+  }
+
+  return kNoError;
+}
+
 int AudioProcessingImpl::AnalyzeReverseStream(const float* const* data,
                                               int samples_per_channel,
-                                              int sample_rate_hz,
+                                              int rev_sample_rate_hz,
                                               ChannelLayout layout) {
   const StreamConfig reverse_config = {
-      sample_rate_hz, ChannelsFromLayout(layout), LayoutHasKeyboard(layout),
+      rev_sample_rate_hz, ChannelsFromLayout(layout), LayoutHasKeyboard(layout),
   };
   if (samples_per_channel != reverse_config.num_frames()) {
     return kBadDataLengthError;
@@ -694,6 +717,17 @@ int AudioProcessingImpl::AnalyzeReverseStream(const float* const* data,
   return AnalyzeReverseStream(data, reverse_config);
 }
 
+int AudioProcessingImpl::ProcessReverseStream(
+    float* const* data,
+    const StreamConfig& reverse_config) {
+  RETURN_ON_ERR(AnalyzeReverseStream(data, reverse_config));
+  if (is_rev_processed()) {
+    render_audio_->CopyTo(api_format_.reverse_stream(), data);
+  }
+
+  return kNoError;
+}
+
 int AudioProcessingImpl::AnalyzeReverseStream(
     const float* const* data,
     const StreamConfig& reverse_config) {
@@ -726,7 +760,16 @@ int AudioProcessingImpl::AnalyzeReverseStream(
 #endif
 
   render_audio_->CopyFrom(data, api_format_.reverse_stream());
-  return AnalyzeReverseStreamLocked();
+  return ProcessReverseStreamLocked();
+}
+
+int AudioProcessingImpl::ProcessReverseStream(AudioFrame* frame) {
+  RETURN_ON_ERR(AnalyzeReverseStream(frame));
+  if (is_rev_processed()) {
+    render_audio_->InterleaveTo(frame, true);
+  }
+
+  return kNoError;
 }
 
 int AudioProcessingImpl::AnalyzeReverseStream(AudioFrame* frame) {
@@ -770,23 +813,32 @@ int AudioProcessingImpl::AnalyzeReverseStream(AudioFrame* frame) {
     RETURN_ON_ERR(WriteMessageToDebugFile());
   }
 #endif
-
   render_audio_->DeinterleaveFrom(frame);
-  return AnalyzeReverseStreamLocked();
+  return ProcessReverseStreamLocked();
 }
 
-int AudioProcessingImpl::AnalyzeReverseStreamLocked() {
+int AudioProcessingImpl::ProcessReverseStreamLocked() {
   AudioBuffer* ra = render_audio_.get();  // For brevity.
   if (rev_proc_format_.sample_rate_hz() == kSampleRate32kHz) {
     ra->SplitIntoFrequencyBands();
   }
 
+  if (intelligibility_enabled_) {
+    intelligibility_enhancer_->ProcessRenderAudio(
+        ra->split_channels_f(kBand0To8kHz), split_rate_, ra->num_channels());
+  }
+
   RETURN_ON_ERR(echo_cancellation_->ProcessRenderAudio(ra));
   RETURN_ON_ERR(echo_control_mobile_->ProcessRenderAudio(ra));
   if (!use_new_agc_) {
     RETURN_ON_ERR(gain_control_->ProcessRenderAudio(ra));
   }
 
+  if (rev_proc_format_.sample_rate_hz() == kSampleRate32kHz &&
+      is_rev_processed()) {
+    ra->MergeFrequencyBands();
+  }
+
   return kNoError;
 }
 
@@ -1003,6 +1055,10 @@ bool AudioProcessingImpl::analysis_needed(bool is_data_processed) const {
   return false;
 }
 
+bool AudioProcessingImpl::is_rev_processed() const {
+  return intelligibility_enabled_ && intelligibility_enhancer_->active();
+}
+
 void AudioProcessingImpl::InitializeExperimentalAgc() {
   if (use_new_agc_) {
     if (!agc_manager_.get()) {
@@ -1035,6 +1091,16 @@ void AudioProcessingImpl::InitializeBeamformer() {
   }
 }
 
+void AudioProcessingImpl::InitializeIntelligibility() {
+  if (intelligibility_enabled_) {
+    IntelligibilityEnhancer::Config config;
+    config.sample_rate_hz = split_rate_;
+    config.num_capture_channels = capture_audio_->num_channels();
+    config.num_render_channels = render_audio_->num_channels();
+    intelligibility_enhancer_.reset(new IntelligibilityEnhancer(config));
+  }
+}
+
 void AudioProcessingImpl::MaybeUpdateHistograms() {
   static const int kMinDiffDelayMs = 60;