Pull the Voice Activity Detector out from the AGC

webrtc/modules/audio_processing/vad/voice_activity_detector_unittest.cc

Index: webrtc/modules/audio_processing/vad/voice_activity_detector_unittest.cc
diff --git a/webrtc/modules/audio_processing/vad/voice_activity_detector_unittest.cc b/webrtc/modules/audio_processing/vad/voice_activity_detector_unittest.cc
new file mode 100644
index 0000000000000000000000000000000000000000..c380ddccd415b90f29e9b839eb6cae66f6ab2ca0
--- /dev/null
+++ b/webrtc/modules/audio_processing/vad/voice_activity_detector_unittest.cc
@@ -0,0 +1,93 @@
+/*
+ *  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 "webrtc/modules/audio_processing/vad/voice_activity_detector.h"
+
+#include <algorithm>
+
+#include "testing/gtest/include/gtest/gtest.h"
+#include "webrtc/test/testsupport/fileutils.h"
+
+namespace webrtc {
+namespace {
+
+const double kDefaultVoiceValue = 1.0;
+const float kMeanSpeechProbability = 0.3f;
+const float kMaxNoiseProbability = 0.05f;

[Andrew MacDonald, 2015/06/17 04:14:57]

Move these three to the tests where they're used.

And just curious, do you actually see values above 0.0?

[aluebs-webrtc, 2015/06/17 17:22:03]

Done.
And yes, there are a few (2 or 3) values which come up to 0.04 or something.

+const size_t kNumChunks = 100u;
+const size_t kNumChunksPerIsacBlock = 3;
+
+void GenerateNoise(int16_t* data, size_t length) {
+  for (size_t i = 0; i < length; ++i) {
+    data[i] = std::rand();

[Andrew MacDonald, 2015/06/17 04:14:57]

std::rand returns between 0 and RAND_MAX. I guess this will work because it
wraps into some random place. Might want to add a comment though.

[aluebs-webrtc, 2015/06/17 17:22:04]

Yes, I know, but I assumed the wrapping would still give me noise. Added
comment.

+  }
+}
+
+}  // namespace
+
+TEST(VoiceActivityDetectorTest, ConstructorSetsDefaultValues) {
+  VoiceActivityDetector vad;
+
+  std::vector<double> p = vad.chunkwise_voice_probabilities();
+  std::vector<double> rms = vad.chunkwise_rms();
+
+  EXPECT_EQ(p.size(), 0u);
+  EXPECT_EQ(rms.size(), 0u);
+
+  EXPECT_DOUBLE_EQ(vad.last_voice_probability(), kDefaultVoiceValue);

[Andrew MacDonald, 2015/06/17 04:14:57]

EXPECT_FLOAT_EQ

[aluebs-webrtc, 2015/06/17 17:22:03]

I forgot to change this after the double->float change. Thanks for catching
that.

+}
+
+TEST(VoiceActivityDetectorTest, SpeechHasHighVoiceProbabilities) {
+  VoiceActivityDetector vad;
+
+  int16_t data[kLength10Ms];
+  float mean_probability = 0.f;
+
+  FILE* pcm_file =
+      fopen(test::ResourcePath("audio_processing/agc/agc_audio", "pcm").c_str(),

[Andrew MacDonald, 2015/06/17 04:14:57]

What is actually in this file? Is it mostly speech? I'm a bit surprised you
can't make the threshold higher.

[aluebs-webrtc, 2015/06/17 17:22:03]

Yes, there is some utterances with some silence in between. Because of the
silences, noise and non-voiced speech I am actually not surprised that the
threshold is that low. It actually is more like 0.39, but I used 0.3 to be more
robust. Maybe I can tweak that a little more, WDYT? Also, I could measure
max_probability instead of mean_probability, which will be definitively higher,
but I think this gives a better picture. WDYT?

+            "rb");
+  ASSERT_TRUE(pcm_file != NULL);
+
+  size_t num_chunks = 0;
+  while (fread(data, sizeof(*data), kLength10Ms, pcm_file) == kLength10Ms) {
+    vad.ProcessChunk(data, kLength10Ms, kSampleRateHz);
+
+    mean_probability += vad.last_voice_probability();
+
+    ++num_chunks;
+  }
+
+  mean_probability /= num_chunks;
+
+  EXPECT_GT(mean_probability, kMeanSpeechProbability);
+}
+
+TEST(VoiceActivityDetectorTest, NoiseHasLowVoiceProbabilities) {
+  VoiceActivityDetector vad;
+
+  int16_t data[kLength10Ms];

[Andrew MacDonald, 2015/06/17 04:14:56]

nit: Use a vector? Then you don't need to pass the length separately to
GenerateNoise.

[aluebs-webrtc, 2015/06/17 17:22:03]

Done.

+  float max_probability = 0.f;
+
+  std::srand(42);
+
+  for (size_t i = 0; i < kNumChunks; ++i) {
+    GenerateNoise(data, kLength10Ms);
+
+    vad.ProcessChunk(data, kLength10Ms, kSampleRateHz);
+
+    if (i > kNumChunksPerIsacBlock) {

[Andrew MacDonald, 2015/06/17 04:14:57]

Do you need to know about this? Why not just check every chunk?

[aluebs-webrtc, 2015/06/17 17:22:04]

Because before the vad has enough data to process an ISAC block it will return
the default value, 1.f, which would ruin the max_probability value.

[Andrew MacDonald, 2015/06/17 21:15:14]

Ah OK. Perhaps add a comment to that effect.

[aluebs-webrtc, 2015/06/18 00:49:21]

Done.

+      max_probability = std::max(max_probability, vad.last_voice_probability());
+    }
+  }
+
+  EXPECT_LT(max_probability, kMaxNoiseProbability);
+}
+
+}  // namespace webrtc