Adding audio RepetitionDetector in AudioProcessingModule.

webrtc/modules/audio_processing/repetition_detector.h

+/*
+ *  Copyright (c) 2015 The WebRTC project authors. All Rights Reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS.  All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef WEBRTC_MODULES_AUDIO_PROCESSING_REPETITION_DETECTOR_H_
+#define WEBRTC_MODULES_AUDIO_PROCESSING_REPETITION_DETECTOR_H_
+
+#include <vector>
+
+#include "webrtc/base/constructormagic.h"
+#include "webrtc/typedefs.h"
+
+namespace webrtc {
+
+class RepetitionDetector {
+ public:
+  RepetitionDetector();
+  virtual ~RepetitionDetector();
+
+  struct Pattern {
+    int id_;
+    // All followings are in samples-per-channel, and are independent of sample
+    // rate, since repetition patterns are supposedly independent of sample

[peah-webrtc, 2015/08/12 11:02:54]

I'm not sure that the repetition pattern length is independent of the sample
rate. Is that the case?

[minyue-webrtc, 2015/08/12 11:32:14]

Right. I had this in my mind when I wrote this, and planned to consult Chromium
people on this. I wrote it out anyway knowing that the change will be small if
we change the unit to a time unit.

Now I've known that it is better to use time unit and a change is at hand.

+    // rate.
+    int look_back_;
+    int look_back_range_;
+    int length_;
+    int length_range_;
+  };
+
+  // Detect repetition given audio samples. When multichannel, samples are
+  // interleaved.
+  void Detect(const int16_t* data, int samples_per_channel,

[Andrew MacDonald, 2015/08/11 20:50:42]

float is preferred now.

[minyue-webrtc, 2015/08/12 11:32:14]

Oh. We depend strongly on the assumption that the repetition consists of bit
exact replica. Is that then too strict an assumption?

[hlundin-webrtc, 2015/08/12 14:12:04]

Use size_t for samples_per_channel and num_channels.

[Andrew MacDonald, 2015/08/12 20:25:32]

Not necessarily, but the Chromium audio pipeline is float, so you gain nothing
by making this int16.

+              int num_channels);
+
+ protected:
+  void RegisterRepititionPatterns(const Pattern* patterns,

[hlundin-webrtc, 2015/08/12 14:12:04]

Why not use some std container class instead of ol' array-and-length?

+                                          int num_patterns);

[hlundin-webrtc, 2015/08/12 14:12:04]

Align.

+  void ClearRepititionPatterns();
+  virtual void ReportRepetition(int id) { }
+
+ private:
+  struct State {
+    State(int id, int look_back, int length, int length_range)
+        : id_(id),
+          look_back_(look_back),
+          length_(length),
+          length_range_(length_range),
+          count_(0) {
+    }
+    int id_;
+    int look_back_;

[hlundin-webrtc, 2015/08/12 14:12:04]

I'd prefer size_t for the length and count variables.

[minyue-webrtc, 2015/08/28 14:27:00]

Now measure in millisecond.

But other counters are made to size_t

+    int length_;
+    int length_range_;
+    int count_;
+  };
+
+  void CreateBuffer();
+  void DeleteBuffer();
+
+  void Reset(int new_num_channels);
+
+  std::vector<State> states_;
+
+  int16_t** audio_buffer_;  // Ring buffers to store input audio channels.

[Andrew MacDonald, 2015/08/11 20:50:42]

Use vectors for this. (Or consider AudioRingBuffer if you want something
full-featured.)

[minyue-webrtc, 2015/08/12 11:32:14]

Acknowledged.

+  int num_channels_;  // Number of audio channels in buffer.
+  int buffer_length_samples_;  // Total samples in buffer for each channel.
+  int buffer_end_;  // Pointer to the end of audio buffer.
+
+  DISALLOW_COPY_AND_ASSIGN(RepetitionDetector);
+};
+
+}  // namespace webrtc
+
+#endif  // WEBRTC_MODULES_AUDIO_PROCESSING_REPETITION_DETECTOR_H_