Bitexactness test for the noise suppressor

webrtc/modules/audio_processing/noise_suppression_bitexactness_unittest.cc

Index: webrtc/modules/audio_processing/noise_suppression_bitexactness_unittest.cc
diff --git a/webrtc/modules/audio_processing/noise_suppression_bitexactness_unittest.cc b/webrtc/modules/audio_processing/noise_suppression_bitexactness_unittest.cc
new file mode 100644
index 0000000000000000000000000000000000000000..1ec36fdcb6f5f2f7624a3a5ce03420d9e0ba19e1
--- /dev/null
+++ b/webrtc/modules/audio_processing/noise_suppression_bitexactness_unittest.cc
@@ -0,0 +1,227 @@
+/*
+ *  Copyright (c) 2016 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 <vector>
+
+#include "testing/gtest/include/gtest/gtest.h"
+#include "webrtc/base/array_view.h"
+#include "webrtc/modules/audio_processing/audio_buffer.h"
+#include "webrtc/modules/audio_processing/noise_suppression_impl.h"
+#include "webrtc/modules/audio_processing/test/audio_buffer_tools.h"
+#include "webrtc/modules/audio_processing/test/bitexactness_tools.h"
+
+namespace webrtc {
+namespace {
+
+const int kNumFramesToProcess = 1000;
+
+// Process one frame of data and produce the output.
+void ProcessOneFrame(int sample_rate_hz,
+                     AudioBuffer* capture_buffer,
+                     NoiseSuppressionImpl* noise_suppressor) {
+  if (sample_rate_hz > AudioProcessing::kSampleRate16kHz) {
+    capture_buffer->SplitIntoFrequencyBands();
+  }
+
+  noise_suppressor->AnalyzeCaptureAudio(capture_buffer);
+  noise_suppressor->ProcessCaptureAudio(capture_buffer);
+
+  if (sample_rate_hz > AudioProcessing::kSampleRate16kHz) {
+    capture_buffer->MergeFrequencyBands();

[hlundin-webrtc, 2016/03/17 14:14:40]

I think I asked this in another CL, but what's the purpose of the merging in
this test? Are you looking at the frame after it is processed above?

[peah-webrtc, 2016/03/17 22:26:02]

The purpose of the merging is that I want to verify the output after the merge.
The subcomponents in APM operate in 16 kHz but extends their processing to
higher frequencies than 8 kHz for 32-48 kHz. Therefore, in order to be able to
verify bitexactness for 32-48 kHz the bands need to be merged again. This is the
same as happens in APM.

Your other comment related to LevelEstimator which is an exception in that it
actually operates on the full band. Therefore, in that CL I removed the band
split/merge. In this CL it is needed though.

Note that after I have processed a the desired number of frames the output is
extracted in RunBitexactness().

[hlundin-webrtc, 2016/03/18 07:53:35]

Acknowledged.

+  }
+}
+
+// Processes a specified amount of frames, verifies the results and reports
+// any errors.
+void RunBitexactnessTest(int sample_rate_hz,
+                         size_t num_channels,
+                         NoiseSuppressionImpl::Level level,
+                         float speech_probability_reference,
+                         rtc::ArrayView<const float> noise_estimate_reference,
+                         rtc::ArrayView<const float> output_reference) {
+  rtc::CriticalSection crit_capture;
+  NoiseSuppressionImpl noise_suppressor(&crit_capture);
+  noise_suppressor.Initialize(num_channels, sample_rate_hz);
+  noise_suppressor.Enable(true);
+  noise_suppressor.set_level(level);
+
+  int samples_per_channel = rtc::CheckedDivExact(sample_rate_hz, 100);
+  const StreamConfig capture_config(sample_rate_hz, num_channels, false);
+  AudioBuffer capture_buffer(
+      capture_config.num_frames(), capture_config.num_channels(),
+      capture_config.num_frames(), capture_config.num_channels(),
+      capture_config.num_frames());
+  test::InputAudioFile capture_file(
+      test::GetApmCaptureTestVectorFileName(sample_rate_hz));
+  std::vector<float> capture_input(samples_per_channel * num_channels);
+  for (size_t frame_no = 0; frame_no < kNumFramesToProcess; ++frame_no) {
+    ReadFloatSamplesFromStereoFile(samples_per_channel, num_channels,
+                                   &capture_file, capture_input);
+
+    test::CopyVectorToAudioBuffer(capture_config, capture_input,
+                                  &capture_buffer);
+
+    ProcessOneFrame(sample_rate_hz, &capture_buffer, &noise_suppressor);
+  }
+
+  // Extract test results.
+  std::vector<float> capture_output;
+  test::ExtractVectorFromAudioBuffer(capture_config, &capture_buffer,
+                                     &capture_output);
+  float speech_probability = noise_suppressor.speech_probability();
+  std::vector<float> noise_estimate = noise_suppressor.NoiseEstimate();
+
+  const float kTolerance = 1.0f / 32768.0f;
+  EXPECT_FLOAT_EQ(speech_probability_reference, speech_probability);
+  EXPECT_TRUE(test::BitExactVector(noise_estimate_reference, noise_estimate,
+                                   kTolerance));
+
+  // Compare the output with the reference. Only the first values of the output
+  // from last frame processed are compared in order not having to specify all
+  // preceeding frames as testvectors. As the algorithm being tested has a
+  // memory, testing only the last frame implicitly also tests the preceeding
+  // frames.
+  EXPECT_TRUE(test::BitExactFrame(
+      capture_config.num_frames(), capture_config.num_channels(),
+      output_reference, capture_output, kTolerance));
+}
+
+}  // namespace
+
+TEST(NoiseSuppresionBitExactnessTest, Mono8kHzLow) {
+#if !defined(WEBRTC_ANDROID)
+  const float kSpeechProbabilityReference = 0.73421317;
+  const float kNoiseEstimateReference[] = {0.035866f, 0.100382f, 0.229889f};
+  const float kOutputReference[] = {0.003263f, 0.004402f, 0.004537f};
+#else
+  const float kSpeechProbabilityReference = 0.73421317;
+  const float kNoiseEstimateReference[] = {0.035866f, 0.100382f, 0.229889f};
+  const float kOutputReference[] = {0.003263f, 0.004402f, 0.004537f};
+#endif
+
+  RunBitexactnessTest(8000, 1, NoiseSuppression::Level::kLow,
+                      kSpeechProbabilityReference, kNoiseEstimateReference,
+                      kOutputReference);
+}
+
+TEST(NoiseSuppresionBitExactnessTest, Mono16kHzLow) {
+#if !defined(WEBRTC_ANDROID)
+  const float kSpeechProbabilityReference = 0.71672988;
+  const float kNoiseEstimateReference[] = {0.065653f, 0.198662f, 0.477870f};
+  const float kOutputReference[] = {0.003574f, 0.004494f, 0.004499f};
+#else
+  const float kSpeechProbabilityReference = 0.71672988;
+  const float kNoiseEstimateReference[] = {0.065653f, 0.198662f, 0.477870f};
+  const float kOutputReference[] = {0.003574f, 0.004494f, 0.004499f};
+#endif
+
+  RunBitexactnessTest(16000, 1, NoiseSuppression::Level::kLow,
+                      kSpeechProbabilityReference, kNoiseEstimateReference,
+                      kOutputReference);
+}
+
+TEST(NoiseSuppresionBitExactnessTest, Mono32kHzLow) {
+#if !defined(WEBRTC_ANDROID)
+  const float kSpeechProbabilityReference = 0.67999554;
+  const float kNoiseEstimateReference[] = {0.065606f, 0.215971f, 0.455931f};
+  const float kOutputReference[] = {0.001221f, 0.001984f, 0.002228f};
+#else
+  const float kSpeechProbabilityReference = 0.67999554;
+  const float kNoiseEstimateReference[] = {0.065606f, 0.215971f, 0.455931f};
+  const float kOutputReference[] = {0.001221f, 0.001984f, 0.002228f};
+#endif
+
+  RunBitexactnessTest(32000, 1, NoiseSuppression::Level::kLow,
+                      kSpeechProbabilityReference, kNoiseEstimateReference,
+                      kOutputReference);
+}
+
+TEST(NoiseSuppresionBitExactnessTest, Mono48kHzLow) {
+#if !defined(WEBRTC_ANDROID)
+  const float kSpeechProbabilityReference = 0.70645678;
+  const float kNoiseEstimateReference[] = {0.066186f, 0.210660f, 0.402548f};
+  const float kOutputReference[] = {-0.013062f, -0.012657f, -0.011934f};
+#else
+  const float kSpeechProbabilityReference = 0.70645678;
+  const float kNoiseEstimateReference[] = {0.066186f, 0.210660f, 0.402548f};
+  const float kOutputReference[] = {-0.013062f, -0.012657f, -0.011934f};
+#endif
+
+  RunBitexactnessTest(48000, 1, NoiseSuppression::Level::kLow,
+                      kSpeechProbabilityReference, kNoiseEstimateReference,
+                      kOutputReference);
+}
+
+TEST(NoiseSuppresionBitExactnessTest, Stereo16kHzLow) {
+#if !defined(WEBRTC_ANDROID)
+  const float kSpeechProbabilityReference = 0.67230678;
+  const float kNoiseEstimateReference[] = {0.298195f, 0.345745f, 0.320528f};
+  const float kOutputReference[] = {-0.011459f, -0.008110f, -0.012728f,
+                                    -0.002399f, 0.001018f,  -0.003189f};
+#else
+  const float kSpeechProbabilityReference = 0.67230678;
+  const float kNoiseEstimateReference[] = {0.298195f, 0.345745f, 0.320528f};
+  const float kOutputReference[] = {-0.011459f, -0.008110f, -0.012728f,
+                                    -0.002399f, 0.001018f,  -0.003189f};
+#endif
+
+  RunBitexactnessTest(16000, 2, NoiseSuppression::Level::kLow,
+                      kSpeechProbabilityReference, kNoiseEstimateReference,
+                      kOutputReference);
+}
+
+TEST(NoiseSuppresionBitExactnessTest, Mono16kHzModerate) {
+#if !defined(WEBRTC_ANDROID)
+  const float kSpeechProbabilityReference = 0.70897013;
+  const float kNoiseEstimateReference[] = {0.066269f, 0.199999f, 0.476885f};
+  const float kOutputReference[] = {0.004513f, 0.005590f, 0.005614f};
+#else
+  const float kSpeechProbabilityReference = 0.70897013;
+  const float kNoiseEstimateReference[] = {0.066269f, 0.199999f, 0.476885f};
+  const float kOutputReference[] = {0.004513f, 0.005590f, 0.005614f};
+#endif
+
+  RunBitexactnessTest(16000, 1, NoiseSuppression::Level::kModerate,
+                      kSpeechProbabilityReference, kNoiseEstimateReference,
+                      kOutputReference);
+}
+
+TEST(NoiseSuppresionBitExactnessTest, Mono16kHzHigh) {
+#if !defined(WEBRTC_ANDROID)
+  const float kSpeechProbabilityReference = 0.70106733;
+  const float kNoiseEstimateReference[] = {0.067901f, 0.204835f, 0.481723f};
+  const float kOutputReference[] = {0.004394f, 0.005406f, 0.005416f};
+#else
+  const float kSpeechProbabilityReference = 0.70106733;
+  const float kNoiseEstimateReference[] = {0.067901f, 0.204835f, 0.481723f};
+  const float kOutputReference[] = {0.004394f, 0.005406f, 0.005416f};
+#endif
+
+  RunBitexactnessTest(16000, 1, NoiseSuppression::Level::kHigh,
+                      kSpeechProbabilityReference, kNoiseEstimateReference,
+                      kOutputReference);
+}
+
+TEST(NoiseSuppresionBitExactnessTest, Mono16kHzVeryHigh) {
+#if !defined(WEBRTC_ANDROID)
+  const float kSpeechProbabilityReference = 0.70281971;
+  const float kNoiseEstimateReference[] = {0.068797f, 0.205191f, 0.481312f};
+  const float kOutputReference[] = {0.004321f, 0.005247f, 0.005263f};
+#else
+  const float kSpeechProbabilityReference = 0.70281971;
+  const float kNoiseEstimateReference[] = {0.068797f, 0.205191f, 0.481312f};
+  const float kOutputReference[] = {0.004321f, 0.005247f, 0.005263f};
+#endif
+
+  RunBitexactnessTest(16000, 1, NoiseSuppression::Level::kVeryHigh,
+                      kSpeechProbabilityReference, kNoiseEstimateReference,
+                      kOutputReference);
+}
+
+}  // namespace webrtc