Adding audio RepetitionDetector in AudioProcessingModule.

webrtc/modules/audio_processing/repetition_detector_unittest.cc

+/*
+ *  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.
+ */
+
+#include <map>
+
+#include "testing/gtest/include/gtest/gtest.h"
+#include "webrtc/base/arraysize.h"
+#include "webrtc/base/scoped_ptr.h"
+#include "webrtc/modules/audio_processing/repetition_detector.h"
+#include "webrtc/modules/remote_bitrate_estimator/test/random.h"
+
+namespace webrtc {
+
+class RepetitionDetectorForTest : public RepetitionDetector {
+ public:
+  int GetCount(int id) {
+    auto it = counters_.find(id);
+    if (it == counters_.end()) {
+      return 0;
+    }
+    return counters_[id];
+  }
+
+  void ResetCounters() {
+    for (auto& item : counters_) {
+      item.second = 0;
+    }
+  }
+
+  void ResetRepetitionPattern(const RepetitionDetector::Pattern* patterns,
+                              size_t num_patterns) {
+    states_.clear();
+    RegisterRepetitionPatterns(patterns, num_patterns);
+  }
+
+  void set_max_frames(size_t max_frames) { max_frames_ = max_frames; }
+
+ private:
+  void ReportRepetition(int id) override {
+    auto it = counters_.find(id);
+    if (it == counters_.end()) {
+      counters_[id] = 0;
+    }
+    counters_[id]++;
+  }
+
+  std::map<int, size_t> counters_;
+};
+
+class RepetitionDetectorTest : public ::testing::Test {
+ protected:
+  struct ExpectedCount {
+    int id_;
+    int count_;
+  };
+
+  // Verify if the counts on the repetition patterns match expectation after
+  // injecting a signal. No reset on the counters
+  void Verify(const ExpectedCount* expected_counts, size_t num_patterns,
+              const float* tester, size_t num_frames,
+              int sample_rate_hz, size_t channels = 1) {
+    detector_.Detect(tester, num_frames, channels, sample_rate_hz);
+    int id;
+    for (size_t idx = 0; idx < num_patterns; idx++) {
+      id = expected_counts[idx].id_;
+      EXPECT_EQ(expected_counts[idx].count_, detector_.GetCount(id)) <<
+          "Repetition #" << id << " counted wrong.";
+    }
+  }
+
+  void VerifyStereo(const ExpectedCount* expected_counts, size_t num_patterns,
+                    const float* tester, size_t num_frames,
+                    int sample_rate_hz) {
+    const size_t kNumChannels = 2;
+
+    // Get memory to store interleaved stereo.
+    rtc::scoped_ptr<float[]> tester_stereo(
+        new float[num_frames * kNumChannels]);
+
+    for (size_t idx = 0; idx < num_frames; ++idx, ++tester) {
+      for (size_t channel = 0; channel < kNumChannels; ++channel) {
+        tester_stereo[idx * kNumChannels + channel] = *tester;
+      }
+    }
+
+    Verify(expected_counts, num_patterns, tester_stereo.get(),
+           num_frames, sample_rate_hz, kNumChannels);
+  }
+
+  void ResetRepetitionPattern(const RepetitionDetector::Pattern* patterns,
+                              size_t num_patterns) {
+    detector_.ResetRepetitionPattern(patterns, num_patterns);
+  }
+
+  void SetMaxFrames(size_t max_frames) {
+    detector_.set_max_frames(max_frames);
+  }
+
+  void ResetCounters() {
+    detector_.ResetCounters();
+  }
+
+ private:
+  RepetitionDetectorForTest detector_;
+};
+
+TEST_F(RepetitionDetectorTest, Basic) {
+  // To make the test signal most obvious, we choose a special sample rate.
+  const int kSampleRateHz = 1000;
+
+  const RepetitionDetector::Pattern kRepetitionPatterns[] = {
+      // id, look_back_ms, min_length_ms
+      {0, 3, 3}
+  };
+  const float kTestSignal[] = {1, 2, 3, 1, 2, 3};
+  const ExpectedCount kExpectedCounts_1[] = {
+      {0, 1}
+  };
+  const ExpectedCount kExpectedCounts_2[] = {
+      {0, 1}
+  };
+
+  ResetRepetitionPattern(kRepetitionPatterns, arraysize(kRepetitionPatterns));
+  Verify(kExpectedCounts_1, arraysize(kExpectedCounts_1), kTestSignal,
+         arraysize(kTestSignal), kSampleRateHz);
+  Verify(kExpectedCounts_2, arraysize(kExpectedCounts_2), kTestSignal,
+         arraysize(kTestSignal), kSampleRateHz);
+  ResetCounters();
+
+  VerifyStereo(kExpectedCounts_1, arraysize(kExpectedCounts_1), kTestSignal,
+               arraysize(kTestSignal), kSampleRateHz);
+  VerifyStereo(kExpectedCounts_2, arraysize(kExpectedCounts_2), kTestSignal,
+               arraysize(kTestSignal), kSampleRateHz);
+}
+
+TEST_F(RepetitionDetectorTest, StereoOutOfSync) {
+  // To make the test signal most obvious, we choose a special sample rate.
+  const int kSampleRateHz = 1000;
+
+  const RepetitionDetector::Pattern kRepetitionPatterns[] = {
+      // id, look_back_ms, min_length_ms
+      {0, 3, 3}
+  };
+  const float kTestSignal[] = {
+      1, 1,
+      2, 2,
+      3, 3,
+      1, 1,
+      2, 2,
+      3, 1};
+  const ExpectedCount kExpectedCounts[] = {
+      {0, 0}
+  };
+
+  ResetRepetitionPattern(kRepetitionPatterns, arraysize(kRepetitionPatterns));
+  Verify(kExpectedCounts, arraysize(kExpectedCounts), kTestSignal,
+         arraysize(kTestSignal) / 2, kSampleRateHz, 2);
+}
+
+TEST_F(RepetitionDetectorTest, IncompletePattern) {
+  // To make the test signal most obvious, we choose a special sample rate.
+  const int kSampleRateHz = 1000;
+
+  const RepetitionDetector::Pattern kRepetitionPatterns[] = {
+      // id, look_back_ms, min_length_ms
+      {0, 3, 3},
+  };
+  const float kTestSignal[] = {1, 2, 1, 2, 3, 1, 2, 3};
+  const ExpectedCount kExpectedCounts[] = {
+      {0, 1},
+  };
+
+  ResetRepetitionPattern(kRepetitionPatterns, arraysize(kRepetitionPatterns));
+  Verify(kExpectedCounts, arraysize(kExpectedCounts), kTestSignal,
+         arraysize(kTestSignal), kSampleRateHz);
+  ResetCounters();
+  VerifyStereo(kExpectedCounts, arraysize(kExpectedCounts), kTestSignal,
+               arraysize(kTestSignal), kSampleRateHz);
+}
+
+TEST_F(RepetitionDetectorTest, PatternLongerThanFrame) {
+  // To make the test signal most obvious, we choose a special sample rate.
+  const int kSampleRateHz = 1000;
+
+  const RepetitionDetector::Pattern kRepetitionPatterns[] = {
+      // id, look_back_ms, min_length_ms
+      {0, 6, 6},
+  };
+  const float kTestSignal_1[] = {1, 2, 3, 4, 5};
+  const float kTestSignal_2[] = {6, 1, 2, 3, 4, 5, 6};
+  const ExpectedCount kExpectedCounts_1[] = {
+      {0, 0},
+  };
+  const ExpectedCount kExpectedCounts_2[] = {
+      {0, 1},
+  };
+
+  ResetRepetitionPattern(kRepetitionPatterns, arraysize(kRepetitionPatterns));
+  Verify(kExpectedCounts_1, arraysize(kExpectedCounts_1), kTestSignal_1,
+         arraysize(kTestSignal_1), kSampleRateHz);
+  Verify(kExpectedCounts_2, arraysize(kExpectedCounts_2), kTestSignal_2,
+         arraysize(kTestSignal_2), kSampleRateHz);
+  ResetCounters();
+  VerifyStereo(kExpectedCounts_1, arraysize(kExpectedCounts_1), kTestSignal_1,
+               arraysize(kTestSignal_1), kSampleRateHz);
+  VerifyStereo(kExpectedCounts_2, arraysize(kExpectedCounts_2), kTestSignal_2,
+               arraysize(kTestSignal_2), kSampleRateHz);
+}
+
+TEST_F(RepetitionDetectorTest, TwoPatterns) {
+  // To make the test signal most obvious, we choose a special sample rate.
+  const int kSampleRateHz = 1000;
+
+  const RepetitionDetector::Pattern kRepetitionPatterns[] = {
+      // id, look_back_ms, min_length_ms
+      {0, 3, 3},
+      {1, 4, 4},
+  };
+  const float kTestSignal[] = {1, 2, 3, 1, 2, 3, 4, 1, 2, 3, 4};
+  const ExpectedCount kExpectedCounts[] = {
+      // 1,2,3 belongs to both patterns.
+      {0, 1},
+      {1, 1}
+  };
+
+  ResetRepetitionPattern(kRepetitionPatterns, arraysize(kRepetitionPatterns));
+  Verify(kExpectedCounts, arraysize(kExpectedCounts), kTestSignal,
+         arraysize(kTestSignal), kSampleRateHz);
+  ResetCounters();
+  VerifyStereo(kExpectedCounts, arraysize(kExpectedCounts), kTestSignal,
+               arraysize(kTestSignal), kSampleRateHz);
+}
+
+TEST_F(RepetitionDetectorTest, MaxFramesShorterThanInput) {
+  // To make the test signal most obvious, we choose a special sample rate.
+  const int kSampleRateHz = 1000;
+
+  const RepetitionDetector::Pattern kRepetitionPatterns[] = {
+      // id, look_back_ms, min_length_ms
+      {0, 3, 3},
+      {1, 4, 4},
+  };
+  const float kTestSignal[] = {1, 2, 3, 1, 2, 3, 4, 1, 2, 3, 4};
+  const ExpectedCount kExpectedCounts[] = {
+      // 1,2,3 belongs to both patterns.
+      {0, 1},
+      {1, 1}
+  };
+
+  // length of kTestSignal is 11 but I set maximum frames to be 2. The detection
+  // should still work.
+  SetMaxFrames(2);
+  ResetRepetitionPattern(kRepetitionPatterns, arraysize(kRepetitionPatterns));
+  Verify(kExpectedCounts, arraysize(kExpectedCounts), kTestSignal,
+         arraysize(kTestSignal), kSampleRateHz);
+  ResetCounters();
+  VerifyStereo(kExpectedCounts, arraysize(kExpectedCounts), kTestSignal,
+               arraysize(kTestSignal), kSampleRateHz);
+}
+
+TEST_F(RepetitionDetectorTest, NestedPatterns) {
+  // To make the test signal most obvious, we choose a special sample rate.
+  const int kSampleRateHz = 1000;
+
+  const RepetitionDetector::Pattern kRepetitionPatterns[] = {
+      // id, look_back_ms, min_length_ms
+      {0, 3, 3},
+      {1, 6, 6},  // When a triplet repeated 3 times, this is triggered.
+  };
+  const float kTestSignal[] = {1, 2, 3, 1, 2, 3};
+  const ExpectedCount kExpectedCounts_1[] = {
+      {0, 1},
+      {1, 0}
+  };
+  const ExpectedCount kExpectedCounts_2[] = {
+      {0, 1},
+      {1, 1}
+  };
+
+  ResetRepetitionPattern(kRepetitionPatterns, arraysize(kRepetitionPatterns));
+  Verify(kExpectedCounts_1, arraysize(kExpectedCounts_1), kTestSignal,
+         arraysize(kTestSignal), kSampleRateHz);
+  Verify(kExpectedCounts_2, arraysize(kExpectedCounts_2), kTestSignal,
+         arraysize(kTestSignal), kSampleRateHz);
+  ResetCounters();
+  VerifyStereo(kExpectedCounts_1, arraysize(kExpectedCounts_1), kTestSignal,
+               arraysize(kTestSignal), kSampleRateHz);
+  VerifyStereo(kExpectedCounts_2, arraysize(kExpectedCounts_2), kTestSignal,
+               arraysize(kTestSignal), kSampleRateHz);
+}
+
+TEST_F(RepetitionDetectorTest, NotFullLengthPattern) {
+  // To make the test signal most obvious, we choose a special sample rate.
+  const int kSampleRateHz = 1000;
+
+  const RepetitionDetector::Pattern kRepetitionPatterns[] = {
+      // id, look_back_ms, min_length_ms
+      {0, 4, 3},
+  };
+  const float kTestSignal[] = {1, 2, 3, -1, 1, 2, 3, -2};
+  const ExpectedCount kExpectedCounts[] = {
+      {0, 1},
+  };
+
+  ResetRepetitionPattern(kRepetitionPatterns, arraysize(kRepetitionPatterns));
+  Verify(kExpectedCounts, arraysize(kExpectedCounts), kTestSignal,
+         arraysize(kTestSignal), kSampleRateHz);
+  ResetCounters();
+  VerifyStereo(kExpectedCounts, arraysize(kExpectedCounts), kTestSignal,
+               arraysize(kTestSignal), kSampleRateHz);
+}
+
+TEST_F(RepetitionDetectorTest, ZerosCountOrNot) {
+  // To make the test signal most obvious, we choose a special sample rate.
+  const int kSampleRateHz = 1000;
+
+  const RepetitionDetector::Pattern kRepetitionPatterns[] = {
+      // id, look_back_ms, min_length_ms
+      {0, 3, 3},
+  };
+  const float kTestSignal_1[] = {0, 0, 0, 0, 0, 0};
+  const float kTestSignal_2[] = {0, 1, 2, 0, 1, 2};
+  const ExpectedCount kExpectedCounts_1[] = {
+      // Full zeros won't count.
+      {0, 0},
+  };
+  const ExpectedCount kExpectedCounts_2[] = {
+      // Partial zero will count.
+      {0, 1},
+  };
+
+  ResetRepetitionPattern(kRepetitionPatterns, arraysize(kRepetitionPatterns));
+  Verify(kExpectedCounts_1, arraysize(kExpectedCounts_1), kTestSignal_1,
+         arraysize(kTestSignal_1), kSampleRateHz);
+  Verify(kExpectedCounts_2, arraysize(kExpectedCounts_2), kTestSignal_2,
+         arraysize(kTestSignal_2), kSampleRateHz);
+  ResetCounters();
+  VerifyStereo(kExpectedCounts_1, arraysize(kExpectedCounts_1), kTestSignal_1,
+               arraysize(kTestSignal_1), kSampleRateHz);
+  VerifyStereo(kExpectedCounts_2, arraysize(kExpectedCounts_2), kTestSignal_2,
+               arraysize(kTestSignal_2), kSampleRateHz);
+}
+
+// Previous tests use short signal to test the detection algorithm, this one
+// tests the built-in pattern in RepetitionDetector.
+TEST_F(RepetitionDetectorTest, BuiltInPattern) {
+  const int kSampleRateHz = 44100;
+  // Let the signal be "*(4ms)-A(13ms)-*(100ms)-A", where * denotes random
+  // samples.
+  const size_t kPreSamples = kSampleRateHz * 4 / 1000;
+  const size_t kRepSamples = kSampleRateHz * 13 / 1000;
+  const size_t kSkipSamples = kSampleRateHz * 100 / 1000;
+  const size_t kSamples = kPreSamples + kRepSamples * 2 + kSkipSamples;
+  float test_signal[kSamples];
+  Random random(0x12345678);
+  size_t idx = 0;
+  for (; idx < kPreSamples + kRepSamples + kSkipSamples; ++idx) {
+    test_signal[idx] = random.Rand();
+  }
+  for (; idx < kSamples; ++idx) {
+    test_signal[idx] = test_signal[idx - kSkipSamples];
+  }
+  const ExpectedCount kExpectedCounts[] = {
+      // Partial zero will count.
+      {0, 0},  // 10 ms look back
+      {1, 1}  // 100 ms look back
+  };
+  Verify(kExpectedCounts, arraysize(kExpectedCounts), test_signal, kSamples,
+         kSampleRateHz);
+}
+
+}  // namespace webrtc