Added first layer of the echo canceller 3 functionality

webrtc/modules/audio_processing/aec3/echo_canceller3.cc

Index: webrtc/modules/audio_processing/aec3/echo_canceller3.cc
diff --git a/webrtc/modules/audio_processing/aec3/echo_canceller3.cc b/webrtc/modules/audio_processing/aec3/echo_canceller3.cc
index e69ccdcbc63fe068a09058e97e32518d6050af05..fa354689c0401c3cec35bfde15eb28565450e457 100644
--- a/webrtc/modules/audio_processing/aec3/echo_canceller3.cc
+++ b/webrtc/modules/audio_processing/aec3/echo_canceller3.cc
@@ -10,35 +10,323 @@
 #include "webrtc/modules/audio_processing/aec3/echo_canceller3.h"
 
 #include "webrtc/base/atomicops.h"
-#include "webrtc/system_wrappers/include/logging.h"
+#include "webrtc/modules/audio_processing/logging/apm_data_dumper.h"
 
 namespace webrtc {
 
+namespace {
+
+bool DetectSaturation(rtc::ArrayView<const float> y) {
+  for (auto y_k : y) {
+    if (y_k >= 32767.0f || y_k <= -32768.0f) {
+      return true;
+    }
+  }
+  return false;
+}
+
+void FillSubFrameView(AudioBuffer* frame,
+                      size_t sub_frame_index,
+                      rtc::ArrayView<rtc::ArrayView<float>> sub_frame_view) {

[aleloi, 2016/12/20 15:55:35]

Can this FillSubFrameView be implemented in terms of the one below with <const
float>?. Also, maybe check that sub_frame_index is in range either here or at
call site.

[peah-webrtc, 2016/12/21 23:13:51]

I changed this to instead used a vector stored on the state. That became better.

Added the checks on sub_frame_index.

Done.

+  for (size_t k = 0; k < frame->num_bands(); ++k) {
+    sub_frame_view[k] = rtc::ArrayView<float>(
+        &frame->split_bands_f(0)[k][sub_frame_index * kSubFrameLength],
+        kSubFrameLength);
+  }
+}
+
+void FillSubFrameView(
+    AudioBuffer* frame,
+    size_t sub_frame_index,
+    rtc::ArrayView<rtc::ArrayView<const float>> sub_frame_view) {
+  for (size_t k = 0; k < frame->num_bands(); ++k) {
+    sub_frame_view[k] = rtc::ArrayView<const float>(
+        &frame->split_bands_f(0)[k][sub_frame_index * kSubFrameLength],
+        kSubFrameLength);
+  }
+}
+
+void FillSubFrameView(
+    const std::vector<std::vector<float>>& frame,
+    size_t sub_frame_index,
+    rtc::ArrayView<rtc::ArrayView<const float>> sub_frame_view) {
+  for (size_t k = 0; k < frame.size(); ++k) {
+    sub_frame_view[k] = rtc::ArrayView<const float>(
+        &frame[k][sub_frame_index * kSubFrameLength], kSubFrameLength);
+  }
+}
+
+void ProcessCaptureFrameContent(AudioBuffer* capture,
+                                bool known_echo_path_change,
+                                bool saturated_microphone_signal,
+                                size_t sub_frame_index,
+                                FrameBlocker* capture_blocker,
+                                BlockFramer* output_framer,
+                                BlockProcessor* block_processor,
+                                std::vector<std::vector<float>>* block) {
+  // Array of ArrayViews.
+  rtc::ArrayView<float> sub_frame_view_data[block->size()];

[hlundin-webrtc, 2016/12/20 15:10:35]

This is not pretty. Sorry...
Is there any way we can improve this code?

[aleloi, 2016/12/21 10:07:45]

Parts of the complexity is here because ArrayViews are created. I think that the
main use case of ArrayView is to hide unsafe pointers to blocks of memory. When
new memory is created, like the array of array views below, ArrayView need not
be the best choice, because we do not need to have any raw pointers to begin
with. Instead maybe use std::vector or std::array (since it's fixed size)?

[peah-webrtc, 2016/12/21 23:13:50]

I changed to using a vector that is cached on the state. That looks better, not
great but at least better. I think that is as far as we come with the current
implementation of AudioBuffer. 

Done.

[peah-webrtc, 2016/12/21 23:13:51]

Agree. I instead solved this by using a vector stored on the state. I don't want
to create the vector for each frame as that includes dynamic allocation, but
storing it on the state avoids that.

PTAL

+  rtc::ArrayView<const float> sub_frame_const_view_data[block->size()];
+  auto sub_frame_view = rtc::ArrayView<rtc::ArrayView<float>>(
+      &sub_frame_view_data[0], block->size());
+  auto sub_frame_const_view = rtc::ArrayView<rtc::ArrayView<const float>>(
+      &sub_frame_const_view_data[0], block->size());
+
+  FillSubFrameView(capture, sub_frame_index, sub_frame_view);

[aleloi, 2016/12/21 10:07:45]

Better to move creation and filling of sub_frame_view below to where it is used,
I think.

[peah-webrtc, 2016/12/21 23:13:51]

Agree.
With the new change, it looks better.
PTAL

+  FillSubFrameView(capture, sub_frame_index, sub_frame_const_view);
+
+  capture_blocker->InsertSubFrameAndExtractBlock(sub_frame_const_view, block);

[aleloi, 2016/12/21 10:07:45]

Is there a type error if a non-const view is passed? If not, maybe we could get
by with less FillSubFrameViews?

[peah-webrtc, 2016/12/21 23:13:50]

Yes, there was a type error. I don't think the automatic const type conversion
works for rtc::ArrayView<rtc::ArrayView<>> types. I did the above mentioned
workaround instead.

PTAL

+  block_processor->ProcessCapture(known_echo_path_change,
+                                  saturated_microphone_signal, block);
+  output_framer->InsertBlockAndExtractSubFrame(*block, sub_frame_view);
+}
+
+void ProcessRemainingCaptureFrameContent(
+    bool known_echo_path_change,
+    bool saturated_microphone_signal,
+    FrameBlocker* capture_blocker,
+    BlockFramer* output_framer,
+    BlockProcessor* block_processor,
+    std::vector<std::vector<float>>* block) {
+  if (!capture_blocker->IsBlockAvailable()) {
+    return;
+  }
+
+  capture_blocker->ExtractBlock(block);
+  block_processor->ProcessCapture(known_echo_path_change,
+                                  saturated_microphone_signal, block);
+  output_framer->InsertBlock(*block);
+}
+
+bool BufferRenderFrameContent(
+    const std::vector<std::vector<float>>& render_frame,
+    size_t sub_frame_index,
+    FrameBlocker* render_blocker,
+    BlockProcessor* block_processor,
+    std::vector<std::vector<float>>* block) {
+  // Array of ArrayViews.
+  rtc::ArrayView<const float> sub_frame_const_view_data[block->size()];
+  auto sub_frame_const_view = rtc::ArrayView<rtc::ArrayView<const float>>(
+      &sub_frame_const_view_data[0], block->size());
+
+  FillSubFrameView(render_frame, sub_frame_index, sub_frame_const_view);
+
+  render_blocker->InsertSubFrameAndExtractBlock(sub_frame_const_view, block);
+  return block_processor->BufferRender(block);
+}
+
+bool BufferRemainingRenderFrameContent(FrameBlocker* render_blocker,
+                                       BlockProcessor* block_processor,
+                                       std::vector<std::vector<float>>* block) {
+  if (!render_blocker->IsBlockAvailable()) {
+    return false;
+  }
+  render_blocker->ExtractBlock(block);
+  return block_processor->BufferRender(block);
+}
+
+void CopyAudioBufferIntoFrame(AudioBuffer* buffer,
+                              size_t num_bands,
+                              size_t frame_length,
+                              std::vector<std::vector<float>>* frame) {
+  RTC_DCHECK_EQ(num_bands, frame->size());
+  for (size_t i = 0; i < num_bands; ++i) {
+    rtc::ArrayView<float> buffer_view(&buffer->split_bands_f(0)[i][0],
+                                      frame_length);
+    std::copy(buffer_view.begin(), buffer_view.end(), (*frame)[i].begin());
+  }
+}
+
+// [B,A] = butter(2,100/4000,'high')
+const CascadedBiQuadFilter::BiQuadCoefficients
+    kHighPassFilterCoefficients_8kHz = {
+        {0.945976856002790, -1.891953712005580, 0.945976856002790},
+        {-1.889033079394525, 0.894874344616636}};
+const int kNumberOfHighPassBiQuads_8kHz = 1;
+
+// [B,A] = butter(2,100/8000,'high')
+const CascadedBiQuadFilter::BiQuadCoefficients
+    kHighPassFilterCoefficients_16kHz = {
+        {0.972613898499844, -1.945227796999688, 0.972613898499844},
+        {-1.944477657767094, 0.945977936232282}};
+const int kNumberOfHighPassBiQuads_16kHz = 1;
+
+static const size_t kRenderTransferQueueSize = 30;
+
+}  // namespace
+
+class EchoCanceller3::RenderWriter {
+ public:
+  RenderWriter(ApmDataDumper* data_dumper,
+               SwapQueue<std::vector<std::vector<float>>,
+                         Aec3RenderQueueItemVerifier>* render_transfer_queue,
+               std::unique_ptr<CascadedBiQuadFilter> render_highpass_filter,
+               int sample_rate_hz,
+               int frame_length,
+               int num_bands);
+  ~RenderWriter();
+  bool Insert(AudioBuffer* render);
+
+ private:
+  ApmDataDumper* data_dumper_;
+  const int sample_rate_hz_;
+  const size_t frame_length_;
+  const int num_bands_;
+  std::unique_ptr<CascadedBiQuadFilter> render_highpass_filter_;
+  std::vector<std::vector<float>> render_queue_input_frame_;
+  SwapQueue<std::vector<std::vector<float>>, Aec3RenderQueueItemVerifier>*
+      render_transfer_queue_;
+  RTC_DISALLOW_IMPLICIT_CONSTRUCTORS(RenderWriter);
+};
+
+EchoCanceller3::RenderWriter::RenderWriter(
+    ApmDataDumper* data_dumper,
+    SwapQueue<std::vector<std::vector<float>>, Aec3RenderQueueItemVerifier>*
+        render_transfer_queue,
+    std::unique_ptr<CascadedBiQuadFilter> render_highpass_filter,
+    int sample_rate_hz,
+    int frame_length,
+    int num_bands)
+    : data_dumper_(data_dumper),
+      sample_rate_hz_(sample_rate_hz),
+      frame_length_(frame_length),
+      num_bands_(num_bands),
+      render_highpass_filter_(std::move(render_highpass_filter)),
+      render_queue_input_frame_(num_bands_,
+                                std::vector<float>(frame_length_, 0.f)),
+      render_transfer_queue_(render_transfer_queue) {
+  RTC_DCHECK(data_dumper);
+}
+
+EchoCanceller3::RenderWriter::~RenderWriter() = default;
+
+bool EchoCanceller3::RenderWriter::Insert(AudioBuffer* input) {
+  RTC_DCHECK_EQ(1, input->num_channels());
+  RTC_DCHECK_EQ(frame_length_, input->num_frames_per_band());
+  data_dumper_->DumpWav("aec3_render_input", frame_length_,
+                        &input->split_bands_f(0)[0][0],
+                        LowestBandRate(sample_rate_hz_), 1);
+
+  CopyAudioBufferIntoFrame(input, num_bands_, frame_length_,
+                           &render_queue_input_frame_);
+
+  if (render_highpass_filter_) {
+    render_highpass_filter_->Process(render_queue_input_frame_[0]);
+  }
+
+  return render_transfer_queue_->Insert(&render_queue_input_frame_);
+}
+
 int EchoCanceller3::instance_count_ = 0;
 
-EchoCanceller3::EchoCanceller3(int sample_rate_hz, bool use_anti_hum_filter) {
-  int band_sample_rate_hz = (sample_rate_hz == 8000 ? sample_rate_hz : 16000);
-  frame_length_ = rtc::CheckedDivExact(band_sample_rate_hz, 100);
+EchoCanceller3::EchoCanceller3(int sample_rate_hz, bool use_highpass_filter)
+    : EchoCanceller3(sample_rate_hz,
+                     use_highpass_filter,
+                     BlockProcessor::Create(sample_rate_hz,
+                                            NumBandsForRate(sample_rate_hz))) {}
+EchoCanceller3::EchoCanceller3(int sample_rate_hz,
+                               bool use_highpass_filter,
+                               BlockProcessor* block_processor)
+    : data_dumper_(new ApmDataDumper(instance_count_)),
+      sample_rate_hz_(sample_rate_hz),
+      num_bands_(NumBandsForRate(sample_rate_hz_)),
+      frame_length_(rtc::CheckedDivExact(LowestBandRate(sample_rate_hz_), 100)),
+      output_framer_(num_bands_),
+      capture_blocker_(num_bands_),
+      render_blocker_(num_bands_),
+      render_transfer_queue_(
+          kRenderTransferQueueSize,
+          std::vector<std::vector<float>>(
+              num_bands_,
+              std::vector<float>(frame_length_, 0.f)),
+          Aec3RenderQueueItemVerifier(num_bands_, frame_length_)),
+      block_processor_(block_processor),
+      render_queue_output_frame_(num_bands_,
+                                 std::vector<float>(frame_length_, 0.f)),
+      block_(num_bands_, std::vector<float>(kBlockSize, 0.f)) {
+  std::unique_ptr<CascadedBiQuadFilter> render_highpass_filter;
+  if (use_highpass_filter) {
+    render_highpass_filter.reset(new CascadedBiQuadFilter(
+        sample_rate_hz_ == 8000 ? kHighPassFilterCoefficients_8kHz
+                                : kHighPassFilterCoefficients_16kHz,
+        sample_rate_hz_ == 8000 ? kNumberOfHighPassBiQuads_8kHz
+                                : kNumberOfHighPassBiQuads_16kHz));
+    capture_highpass_filter_.reset(new CascadedBiQuadFilter(
+        sample_rate_hz_ == 8000 ? kHighPassFilterCoefficients_8kHz
+                                : kHighPassFilterCoefficients_16kHz,
+        sample_rate_hz_ == 8000 ? kNumberOfHighPassBiQuads_8kHz
+                                : kNumberOfHighPassBiQuads_16kHz));
+  } else {
+    render_highpass_filter.reset(nullptr);

[ivoc, 2016/12/21 13:04:05]

Is this really necessary? Isn't it set to nullptr by default?

[peah-webrtc, 2016/12/21 23:13:51]

I'm not sure. The issue was that there was an explicit move of one of these and
there was concern whether one could rely on that then would move a nullptr or
something else. I added this to make sure. Do you know?

I removed one of these, at least, since that one is not needed.

WDYT?

[ivoc, 2016/12/22 13:38:13]

I'm pretty sure that it's not needed, the default value of a unique_ptr is
nullptr, so this would have no effect. Calling .reset(nullptr) is only useful to
delete an object that was previously owned by the unique_ptr.

[peah-webrtc, 2017/01/02 08:45:10]

Thanks! I agree.

Done.

+    capture_highpass_filter_.reset(nullptr);
+  }
+
+  render_writer_.reset(
+      new RenderWriter(data_dumper_.get(), &render_transfer_queue_,
+                       std::move(render_highpass_filter), sample_rate_hz_,
+                       frame_length_, num_bands_));
 
-  LOG(LS_INFO) << "AEC3 created : "
-               << "{ instance_count: " << instance_count_ << "}";
+  RTC_DCHECK_EQ(num_bands_, std::max(sample_rate_hz_, 16000) / 16000);
+  RTC_DCHECK_GE(kMaxNumBands, num_bands_);
   instance_count_ = rtc::AtomicOps::Increment(&instance_count_);
 }
 
 EchoCanceller3::~EchoCanceller3() = default;
 
 bool EchoCanceller3::AnalyzeRender(AudioBuffer* render) {
-  RTC_DCHECK_EQ(1u, render->num_channels());
-  RTC_DCHECK_EQ(frame_length_, render->num_frames_per_band());
-  return true;
+  return render_writer_->Insert(render);
 }
 
-void EchoCanceller3::AnalyzeCapture(AudioBuffer* capture) {}
+void EchoCanceller3::AnalyzeCapture(AudioBuffer* capture) {
+  data_dumper_->DumpWav("aec3_capture_analyze_input", frame_length_,
+                        capture->channels_f()[0], sample_rate_hz_, 1);
+
+  saturated_microphone_signal_ = false;
+  for (size_t k = 0; k < capture->num_channels(); ++k) {
+    saturated_microphone_signal_ |=
+        DetectSaturation(rtc::ArrayView<const float>(capture->channels_f()[k],
+                                                     capture->num_frames()));
+    if (saturated_microphone_signal_) {
+      break;
+    }
+  }
+}
 
 void EchoCanceller3::ProcessCapture(AudioBuffer* capture,
                                     bool known_echo_path_change) {
   RTC_DCHECK_EQ(1u, capture->num_channels());
   RTC_DCHECK_EQ(frame_length_, capture->num_frames_per_band());
+
+  rtc::ArrayView<float> capture_lower_band =
+      rtc::ArrayView<float>(&capture->split_bands_f(0)[0][0], frame_length_);
+
+  data_dumper_->DumpWav("aec3_capture_input", capture_lower_band,
+                        LowestBandRate(sample_rate_hz_), 1);
+
+  bool render_buffer_overrun = EmptyRenderQueue();
+  RTC_DCHECK(!render_buffer_overrun);
+
+  if (capture_highpass_filter_) {
+    capture_highpass_filter_->Process(capture_lower_band);
+  }
+
+  ProcessCaptureFrameContent(capture, known_echo_path_change,
+                             saturated_microphone_signal_, 0, &capture_blocker_,
+                             &output_framer_, block_processor_.get(), &block_);
+
+  if (sample_rate_hz_ != 8000) {
+    ProcessCaptureFrameContent(
+        capture, known_echo_path_change, saturated_microphone_signal_, 1,
+        &capture_blocker_, &output_framer_, block_processor_.get(), &block_);
+  }
+
+  ProcessRemainingCaptureFrameContent(
+      known_echo_path_change, saturated_microphone_signal_, &capture_blocker_,
+      &output_framer_, block_processor_.get(), &block_);
+
+  data_dumper_->DumpWav("aec3_capture_output", frame_length_,
+                        &capture->split_bands_f(0)[0][0],
+                        LowestBandRate(sample_rate_hz_), 1);
 }
 
 std::string EchoCanceller3::ToString(
@@ -54,4 +342,28 @@ bool EchoCanceller3::Validate(
   return true;
 }
 
+bool EchoCanceller3::EmptyRenderQueue() {
+  bool render_buffer_overrun = false;
+  bool frame_to_buffer =
+      render_transfer_queue_.Remove(&render_queue_output_frame_);
+  while (frame_to_buffer) {
+    render_buffer_overrun |= BufferRenderFrameContent(
+        render_queue_output_frame_, 0, &render_blocker_, block_processor_.get(),
+        &block_);
+
+    if (sample_rate_hz_ != 8000) {
+      render_buffer_overrun |= BufferRenderFrameContent(
+          render_queue_output_frame_, 1, &render_blocker_,
+          block_processor_.get(), &block_);
+    }
+
+    render_buffer_overrun |= BufferRemainingRenderFrameContent(
+        &render_blocker_, block_processor_.get(), &block_);
+
+    frame_to_buffer =
+        render_transfer_queue_.Remove(&render_queue_output_frame_);
+  }
+  return render_buffer_overrun;
+}
+
 }  // namespace webrtc