admin/webrtc_m130
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Compensate for the LevelController gain in the IntelligibilityEnhancer

Browse Source 
R=peah@webrtc.org

Review URL: https://codereview.webrtc.org/2112553003 .

Cr-Commit-Position: refs/heads/master@{#13353}
This commit is contained in:
Alejandro Luebs 2016-06-30 15:35:41 -07:00
parent a181c9ad17
commit 5041110b94
4 changed files with 10 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
webrtc/modules/audio_processing/audio_processing_impl.cc
View File

@ -728,8 +728,12 @@ int AudioProcessingImpl::ProcessStreamLocked() {
int gain_db = public_submodules_->gain_control->is_enabled() ?
public_submodules_->gain_control->compression_gain_db() :
0;
float gain = std::pow(10.f, gain_db / 20.f);
gain *= capture_nonlocked_.level_controller_enabled ?
private_submodules_->level_controller->GetLastGain() :
1.f;
public_submodules_->intelligibility_enhancer->SetCaptureNoiseEstimate(
public_submodules_->noise_suppression->NoiseEstimate(), gain_db);
public_submodules_->noise_suppression->NoiseEstimate(), gain);
}
// Ensure that the stream delay was set before the call to the

3
webrtc/modules/audio_processing/intelligibility/intelligibility_enhancer.cc
View File

@ -111,9 +111,8 @@ IntelligibilityEnhancer::IntelligibilityEnhancer(int sample_rate_hz,
}
void IntelligibilityEnhancer::SetCaptureNoiseEstimate(
std::vector<float> noise, int gain_db) {
std::vector<float> noise, float gain) {
RTC_DCHECK_EQ(noise.size(), num_noise_bins_);
const float gain = std::pow(10.f, gain_db / 20.f);
for (auto& bin : noise) {
bin *= gain;
}

2
webrtc/modules/audio_processing/intelligibility/intelligibility_enhancer.h
View File

@ -36,7 +36,7 @@ class IntelligibilityEnhancer : public LappedTransform::Callback {
size_t num_noise_bins);
// Sets the capture noise magnitude spectrum estimate.
void SetCaptureNoiseEstimate(std::vector<float> noise, int gain_db);
void SetCaptureNoiseEstimate(std::vector<float> noise, float gain);
// Reads chunk of speech in time domain and updates with modified signal.
void ProcessRenderAudio(float* const* audio,

7
webrtc/modules/audio_processing/intelligibility/intelligibility_enhancer_unittest.cc
View File

@ -407,18 +407,17 @@ TEST_F(IntelligibilityEnhancerTest, TestSolveForGains) {
}
TEST_F(IntelligibilityEnhancerTest, TestNoiseGainHasExpectedResult) {
const int kGainDB = 6;
const float kGainFactor = std::pow(10.f, kGainDB / 20.f);
const float kGain = 2.f;
const float kTolerance = 0.007f;
std::vector<float> noise(kNumNoiseBins);
std::vector<float> noise_psd(kNumNoiseBins);
std::generate(noise.begin(), noise.end(), float_rand);
for (size_t i = 0; i < kNumNoiseBins; ++i) {
noise_psd[i] = kGainFactor * kGainFactor * noise[i] * noise[i];
noise_psd[i] = kGain * kGain * noise[i] * noise[i];
}
float* clear_cursor = clear_data_.data();
for (size_t i = 0; i < kNumFramesToProcess; ++i) {
enh_->SetCaptureNoiseEstimate(noise, kGainDB);
enh_->SetCaptureNoiseEstimate(noise, kGain);
enh_->ProcessRenderAudio(&clear_cursor, kSampleRate, kNumChannels);
}
const std::vector<float>& estimated_psd =