Add detection for repeated audio in capturing.

content/renderer/media/audio_repetition_detector.cc

+// Copyright 2015 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "content/renderer/media/audio_repetition_detector.h"
+
+#include "base/logging.h"
+#include "base/macros.h"
+#include "base/metrics/histogram_macros.h"
+
+namespace content {
+
+namespace {
+
+const AudioRepetitionDetector::Pattern kRepetitionPatterns[] = {
+    {1, 10, 10},
+    {2, 20, 10},
+    {3, 30, 10},
+    {4, 40, 10},
+    {5, 50, 10},
+    {6, 60, 10},
+    {7, 70, 10},
+    {8, 80, 10},
+    {9, 90, 10},
+    {10, 100, 10},
+    {20, 200, 10},
+};
+
+// This is used for increasing the efficiency of copying data into the buffer.
+// Input longer than |kMaxFrames| won't be a problem, and will be devided into
+// chunks automatically.
+const size_t kMaxFrames = 480;  // 10 ms * 48 kHz
+
+}  // namespace
+
+AudioRepetitionDetector::AudioRepetitionDetector()
+    : max_look_back_ms_(0),
+      sample_rate_(0),
+      buffer_size_frames_(0),
+      buffer_end_index_(0),
+      max_frames_(kMaxFrames) {
+  RegisterRepetitionPatterns(kRepetitionPatterns,
+                             arraysize(kRepetitionPatterns));
+  // May be created in the main render thread and used in the audio threads.
+  thread_checker_.DetachFromThread();
+}
+
+AudioRepetitionDetector::~AudioRepetitionDetector() {
+  DCHECK(thread_checker_.CalledOnValidThread());

[tommi (sloooow) - chröme, 2015/10/02 19:45:31]

looks like we're hitting this in the tests.

[minyue, 2015/10/15 18:35:47]

Yes, a DetachFromThread() at ctor should also mean to remove this check.

+}
+
+
+void AudioRepetitionDetector::Detect(const float* data, size_t num_frames,
+                                     size_t num_channels, int sample_rate) {
+  DCHECK(thread_checker_.CalledOnValidThread());
+  DCHECK(!states_.empty());
+  if (num_channels != num_channels_ || sample_rate != sample_rate_)
+    Reset(num_channels, sample_rate);
+
+  // The maximum number of frames |audio_buffer_| can take in is |max_frames_|.
+  // Therefore, input data with larger frames needs be divided into chunks.
+  const size_t chunk_size = max_frames_ * num_channels;
+  while (num_frames > max_frames_) {
+    Detect(data, max_frames_, num_channels, sample_rate);
+    data += chunk_size;
+    num_frames -= max_frames_;
+  }
+
+  if (num_frames == 0)
+    return;
+
+  AddFramesToBuffer(data, num_frames);
+
+  for (size_t idx = num_frames; idx > 0; --idx, data += num_channels) {
+    for (State* state : states_) {
+      // Look back position depends on the sample rate. It is rounded down to
+      // the closest integer.
+      const size_t look_back_frames =
+          state->look_back_ms() * sample_rate_ / 1000;
+      // Equal(data, offset) checks if |data| equals the audio frame located
+      // |offset| frames from the end of buffer. Now a full frame has been
+      // inserted to the buffer, and thus |offset| should compensate for it.
+      if (Equal(data, look_back_frames + idx)) {
+        if (!state->reported()) {
+          state->Increment(IsZero(data, num_channels));
+          if (state->HasValidReport(sample_rate)) {
+            ReportRepetition(state->id());
+            state->set_reported(true);
+          }
+        }
+      } else {
+        state->Reset();
+      }
+    }
+  }
+}
+
+AudioRepetitionDetector::State::State(const Pattern &pattern)
+    : pattern_(pattern) {
+  Reset();
+}
+
+void AudioRepetitionDetector::State::Increment(bool zero) {
+  if (0 == count_frames_ && zero) {
+    // If a repetition starts with zeros, we enter the all zero mode until
+    // a non zero is found later. The point is that the beginning zeros should
+    // be counted in the length of the repetition as long as the repetition does
+    // not comprise only zeros.
+    all_zero_ = true;
+  }
+  ++count_frames_;
+  if (!zero)
+    all_zero_ = false;
+}
+
+bool AudioRepetitionDetector::State::HasValidReport(int sample_rate) const {
+  return (!all_zero_ && count_frames_ >=
+      static_cast<size_t>(pattern_.min_length_ms * sample_rate / 1000));
+}
+
+void AudioRepetitionDetector::State::Reset() {
+  count_frames_ = 0;
+  all_zero_ = true;
+  reported_ = false;
+}
+
+void AudioRepetitionDetector::RegisterRepetitionPatterns(
+    const Pattern* patterns, size_t num_patterns) {
+  DCHECK(thread_checker_.CalledOnValidThread());
+  Pattern pattern;
+  for (size_t idx = 0; idx < num_patterns; idx++) {
+    pattern = patterns[idx];
+    ids_.push_back(pattern.id);
+    states_.push_back(new State(pattern));
+    if (pattern.look_back_ms > max_look_back_ms_)
+      max_look_back_ms_ = pattern.look_back_ms;
+  }
+}
+
+void AudioRepetitionDetector::Reset(size_t num_channels, int sample_rate) {
+  DCHECK(thread_checker_.CalledOnValidThread());
+  num_channels_ = num_channels;
+  sample_rate_ = sample_rate;
+
+  // |(xxx + 999) / 1000| is an arithmetic way to round up |xxx / 1000|.
+  buffer_size_frames_ =
+      (max_look_back_ms_ * sample_rate_ + 999) / 1000 + max_frames_;
+
+  audio_buffer_.resize(buffer_size_frames_ * num_channels_);
+  for (State* state : states_)
+    state->Reset();
+}
+
+void AudioRepetitionDetector::AddFramesToBuffer(const float* data,
+                                                size_t num_frames) {
+  DCHECK(thread_checker_.CalledOnValidThread());
+  DCHECK_LE(num_frames, buffer_size_frames_);
+  const size_t margin = buffer_size_frames_ - buffer_end_index_;
+  const auto it = audio_buffer_.begin() + buffer_end_index_ * num_channels_;
+  if (num_frames <= margin) {
+    std::copy(data, data + num_frames * num_channels_, it);
+    buffer_end_index_ += num_frames;
+  } else {
+    std::copy(data, data + margin * num_channels_, it);
+    std::copy(data + margin * num_channels_, data + num_frames * num_channels_,
+              audio_buffer_.begin());
+    buffer_end_index_ = num_frames - margin;
+  }
+}
+
+bool AudioRepetitionDetector::Equal(const float* frame,
+                                    int look_back_frames) const {
+  DCHECK(thread_checker_.CalledOnValidThread());
+  const size_t look_back_index =
+      (buffer_end_index_ + buffer_size_frames_ - look_back_frames) %
+      buffer_size_frames_ ;
+  auto it = audio_buffer_.begin() + look_back_index * num_channels_;
+  for (size_t channel = 0; channel < num_channels_; ++channel, ++frame, ++it) {
+    if (*frame != *it)
+      return false;
+  }
+  return true;
+}
+
+bool AudioRepetitionDetector::IsZero(const float* frame,
+                                     size_t num_channels) const {
+  for (size_t channel = 0; channel < num_channels; ++channel, ++frame) {
+    if (*frame != 0)
+      return false;
+  }
+  return true;
+}
+
+void AudioRepetitionDetector::ReportRepetition(int id) {
+  DCHECK(thread_checker_.CalledOnValidThread());
+  UMA_HISTOGRAM_CUSTOM_ENUMERATION(
+      "Media.AudioCapturerRepetition", id, ids_);

[rkaplow, 2015/10/02 14:42:33]

what is the output space for the ids? Normally an enum should have each entry
mapped to a specific output type.

[minyue, 2015/10/02 19:27:51]

Our idea is to use the ID (integer) of some pattern as the histogram bins.
Different from enums, the IDs does not need alias. The current IDs of interest
include {1,2,...10, 20}, 11 to 19 are omitted to make the ids more meaningful.
It is likely that we add more IDs in the future.

I find UMA_HISTOGRAM_CUSTOM_ENUMERATION, which takes a vector<int> as the value
space, to be useful in this scenario. What do you suggest, Robert?

+}
+
+}  // namespace content