1/*
2 *  Copyright (c) 2015 The WebRTC project authors. All Rights Reserved.
3 *
4 *  Use of this source code is governed by a BSD-style license
5 *  that can be found in the LICENSE file in the root of the source
6 *  tree. An additional intellectual property rights grant can be found
7 *  in the file PATENTS.  All contributing project authors may
8 *  be found in the AUTHORS file in the root of the source tree.
9 */
10
11//
12//  Unit tests for intelligibility enhancer.
13//
14
15#include <math.h>
16#include <stdlib.h>
17#include <algorithm>
18#include <vector>
19
20#include "testing/gtest/include/gtest/gtest.h"
21#include "webrtc/base/arraysize.h"
22#include "webrtc/base/scoped_ptr.h"
23#include "webrtc/common_audio/signal_processing/include/signal_processing_library.h"
24#include "webrtc/modules/audio_processing/intelligibility/intelligibility_enhancer.h"
25
26namespace webrtc {
27
28namespace {
29
30// Target output for ERB create test. Generated with matlab.
31const float kTestCenterFreqs[] = {
32    13.169f, 26.965f, 41.423f, 56.577f, 72.461f, 89.113f, 106.57f, 124.88f,
33    144.08f, 164.21f, 185.34f, 207.5f,  230.75f, 255.16f, 280.77f, 307.66f,
34    335.9f,  365.56f, 396.71f, 429.44f, 463.84f, 500.f};
35const float kTestFilterBank[][2] = {{0.055556f, 0.f},
36                                    {0.055556f, 0.f},
37                                    {0.055556f, 0.f},
38                                    {0.055556f, 0.f},
39                                    {0.055556f, 0.f},
40                                    {0.055556f, 0.f},
41                                    {0.055556f, 0.f},
42                                    {0.055556f, 0.f},
43                                    {0.055556f, 0.f},
44                                    {0.055556f, 0.f},
45                                    {0.055556f, 0.f},
46                                    {0.055556f, 0.f},
47                                    {0.055556f, 0.f},
48                                    {0.055556f, 0.f},
49                                    {0.055556f, 0.f},
50                                    {0.055556f, 0.f},
51                                    {0.055556f, 0.f},
52                                    {0.055556f, 0.2f},
53                                    {0, 0.2f},
54                                    {0, 0.2f},
55                                    {0, 0.2f},
56                                    {0, 0.2f}};
57static_assert(arraysize(kTestCenterFreqs) == arraysize(kTestFilterBank),
58              "Test filterbank badly initialized.");
59
60// Target output for gain solving test. Generated with matlab.
61const size_t kTestStartFreq = 12;  // Lowest integral frequency for ERBs.
62const float kTestZeroVar[] = {1.f, 1.f, 1.f, 1.f, 1.f, 1.f, 1.f, 1.f,
63                              1.f, 1.f, 1.f, 0.f, 0.f, 0.f, 0.f, 0.f,
64                              0.f, 0.f, 0.f, 0.f, 0.f, 0.f};
65static_assert(arraysize(kTestCenterFreqs) == arraysize(kTestZeroVar),
66              "Variance test data badly initialized.");
67const float kTestNonZeroVarLambdaTop[] = {
68    1.f,     1.f,     1.f,     1.f,     1.f,     1.f,     1.f,     1.f,
69    1.f,     1.f,     1.f,     0.f,     0.f,     0.0351f, 0.0636f, 0.0863f,
70    0.1037f, 0.1162f, 0.1236f, 0.1251f, 0.1189f, 0.0993f};
71static_assert(arraysize(kTestCenterFreqs) ==
72                  arraysize(kTestNonZeroVarLambdaTop),
73              "Variance test data badly initialized.");
74const float kMaxTestError = 0.005f;
75
76// Enhancer initialization parameters.
77const int kSamples = 2000;
78const int kSampleRate = 1000;
79const int kNumChannels = 1;
80const int kFragmentSize = kSampleRate / 100;
81
82}  // namespace
83
84using std::vector;
85using intelligibility::VarianceArray;
86
87class IntelligibilityEnhancerTest : public ::testing::Test {
88 protected:
89  IntelligibilityEnhancerTest()
90      : clear_data_(kSamples), noise_data_(kSamples), orig_data_(kSamples) {
91    config_.sample_rate_hz = kSampleRate;
92    enh_.reset(new IntelligibilityEnhancer(config_));
93  }
94
95  bool CheckUpdate(VarianceArray::StepType step_type) {
96    config_.sample_rate_hz = kSampleRate;
97    config_.var_type = step_type;
98    enh_.reset(new IntelligibilityEnhancer(config_));
99    float* clear_cursor = &clear_data_[0];
100    float* noise_cursor = &noise_data_[0];
101    for (int i = 0; i < kSamples; i += kFragmentSize) {
102      enh_->AnalyzeCaptureAudio(&noise_cursor, kSampleRate, kNumChannels);
103      enh_->ProcessRenderAudio(&clear_cursor, kSampleRate, kNumChannels);
104      clear_cursor += kFragmentSize;
105      noise_cursor += kFragmentSize;
106    }
107    for (int i = 0; i < kSamples; i++) {
108      if (std::fabs(clear_data_[i] - orig_data_[i]) > kMaxTestError) {
109        return true;
110      }
111    }
112    return false;
113  }
114
115  IntelligibilityEnhancer::Config config_;
116  rtc::scoped_ptr<IntelligibilityEnhancer> enh_;
117  vector<float> clear_data_;
118  vector<float> noise_data_;
119  vector<float> orig_data_;
120};
121
122// For each class of generated data, tests that render stream is
123// updated when it should be for each variance update method.
124TEST_F(IntelligibilityEnhancerTest, TestRenderUpdate) {
125  vector<VarianceArray::StepType> step_types;
126  step_types.push_back(VarianceArray::kStepInfinite);
127  step_types.push_back(VarianceArray::kStepDecaying);
128  step_types.push_back(VarianceArray::kStepWindowed);
129  step_types.push_back(VarianceArray::kStepBlocked);
130  step_types.push_back(VarianceArray::kStepBlockBasedMovingAverage);
131  std::fill(noise_data_.begin(), noise_data_.end(), 0.0f);
132  std::fill(orig_data_.begin(), orig_data_.end(), 0.0f);
133  for (auto step_type : step_types) {
134    std::fill(clear_data_.begin(), clear_data_.end(), 0.0f);
135    EXPECT_FALSE(CheckUpdate(step_type));
136  }
137  std::srand(1);
138  auto float_rand = []() { return std::rand() * 2.f / RAND_MAX - 1; };
139  std::generate(noise_data_.begin(), noise_data_.end(), float_rand);
140  for (auto step_type : step_types) {
141    EXPECT_FALSE(CheckUpdate(step_type));
142  }
143  for (auto step_type : step_types) {
144    std::generate(clear_data_.begin(), clear_data_.end(), float_rand);
145    orig_data_ = clear_data_;
146    EXPECT_TRUE(CheckUpdate(step_type));
147  }
148}
149
150// Tests ERB bank creation, comparing against matlab output.
151TEST_F(IntelligibilityEnhancerTest, TestErbCreation) {
152  ASSERT_EQ(arraysize(kTestCenterFreqs), enh_->bank_size_);
153  for (size_t i = 0; i < enh_->bank_size_; ++i) {
154    EXPECT_NEAR(kTestCenterFreqs[i], enh_->center_freqs_[i], kMaxTestError);
155    ASSERT_EQ(arraysize(kTestFilterBank[0]), enh_->freqs_);
156    for (size_t j = 0; j < enh_->freqs_; ++j) {
157      EXPECT_NEAR(kTestFilterBank[i][j], enh_->filter_bank_[i][j],
158                  kMaxTestError);
159    }
160  }
161}
162
163// Tests analytic solution for optimal gains, comparing
164// against matlab output.
165TEST_F(IntelligibilityEnhancerTest, TestSolveForGains) {
166  ASSERT_EQ(kTestStartFreq, enh_->start_freq_);
167  vector<float> sols(enh_->bank_size_);
168  float lambda = -0.001f;
169  for (size_t i = 0; i < enh_->bank_size_; i++) {
170    enh_->filtered_clear_var_[i] = 0.0f;
171    enh_->filtered_noise_var_[i] = 0.0f;
172    enh_->rho_[i] = 0.02f;
173  }
174  enh_->SolveForGainsGivenLambda(lambda, enh_->start_freq_, &sols[0]);
175  for (size_t i = 0; i < enh_->bank_size_; i++) {
176    EXPECT_NEAR(kTestZeroVar[i], sols[i], kMaxTestError);
177  }
178  for (size_t i = 0; i < enh_->bank_size_; i++) {
179    enh_->filtered_clear_var_[i] = static_cast<float>(i + 1);
180    enh_->filtered_noise_var_[i] = static_cast<float>(enh_->bank_size_ - i);
181  }
182  enh_->SolveForGainsGivenLambda(lambda, enh_->start_freq_, &sols[0]);
183  for (size_t i = 0; i < enh_->bank_size_; i++) {
184    EXPECT_NEAR(kTestNonZeroVarLambdaTop[i], sols[i], kMaxTestError);
185  }
186  lambda = -1.0;
187  enh_->SolveForGainsGivenLambda(lambda, enh_->start_freq_, &sols[0]);
188  for (size_t i = 0; i < enh_->bank_size_; i++) {
189    EXPECT_NEAR(kTestZeroVar[i], sols[i], kMaxTestError);
190  }
191}
192
193}  // namespace webrtc
194