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

AEC3: Change adaptation speed of the matched filter after a delay is found

Browse Source 
This change enables the use of two different adaptation speeds of the
matched filter of the delay estimator of AEC3.

One speed is used when no delay has been found, and one is used after a
reliable delay has been found. The purpose is to use a slower adaptation
speed to reduce the risk of divergence during double-talk without
slowing down the search for the initial delay.

The CL prepares for experimentation by adding field trials for
controlling the two adaptation speeds.

Bug: webrtc:12775
Change-Id: I817a1ab5ded0f78d20de45edcf04c708290173fc
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/219083
Commit-Queue: Gustaf Ullberg <gustaf@webrtc.org>
Reviewed-by: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#34055}
This commit is contained in:
Gustaf Ullberg 2021-05-19 14:26:31 +02:00 committed by WebRTC LUCI CQ
parent 402ceffff1
commit aeb8ce882f
8 changed files with 54 additions and 20 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
api/audio/echo_canceller3_config.h
View File

@ -43,6 +43,7 @@ struct RTC_EXPORT EchoCanceller3Config {
size_t hysteresis_limit_blocks = 1;
size_t fixed_capture_delay_samples = 0;
float delay_estimate_smoothing = 0.7f;
float delay_estimate_smoothing_delay_found = 0.7f;
float delay_candidate_detection_threshold = 0.2f;
struct DelaySelectionThresholds {
int initial;

4
api/audio/echo_canceller3_config_json.cc
View File

@ -191,6 +191,8 @@ void Aec3ConfigFromJsonString(absl::string_view json_string,
&cfg.delay.fixed_capture_delay_samples);
ReadParam(section, "delay_estimate_smoothing",
&cfg.delay.delay_estimate_smoothing);
ReadParam(section, "delay_estimate_smoothing_delay_found",
&cfg.delay.delay_estimate_smoothing_delay_found);
ReadParam(section, "delay_candidate_detection_threshold",
&cfg.delay.delay_candidate_detection_threshold);
@ -425,6 +427,8 @@ std::string Aec3ConfigToJsonString(const EchoCanceller3Config& config) {
<< config.delay.fixed_capture_delay_samples << ",";
ost << "\"delay_estimate_smoothing\": "
<< config.delay.delay_estimate_smoothing << ",";
ost << "\"delay_estimate_smoothing_delay_found\": "
<< config.delay.delay_estimate_smoothing_delay_found << ",";
ost << "\"delay_candidate_detection_threshold\": "
<< config.delay.delay_candidate_detection_threshold << ",";

6
modules/audio_processing/aec3/echo_canceller3.cc
View File

@ -572,6 +572,12 @@ EchoCanceller3Config AdjustConfig(const EchoCanceller3Config& config) {
RetrieveFieldTrialValue("WebRTC-Aec3SuppressorEpStrengthDefaultLenOverride",
-1.f, 1.f, &adjusted_cfg.ep_strength.default_len);
// Field trial-based overrides of individual delay estimator parameters.
RetrieveFieldTrialValue("WebRTC-Aec3DelayEstimateSmoothingOverride", 0.f, 1.f,
&adjusted_cfg.delay.delay_estimate_smoothing);
RetrieveFieldTrialValue(
"WebRTC-Aec3DelayEstimateSmoothingDelayFoundOverride", 0.f, 1.f,
&adjusted_cfg.delay.delay_estimate_smoothing_delay_found);
return adjusted_cfg;
}

4
modules/audio_processing/aec3/echo_path_delay_estimator.cc
View File

@ -42,6 +42,7 @@ EchoPathDelayEstimator::EchoPathDelayEstimator(
? config.render_levels.poor_excitation_render_limit_ds8
: config.render_levels.poor_excitation_render_limit,
config.delay.delay_estimate_smoothing,
config.delay.delay_estimate_smoothing_delay_found,
config.delay.delay_candidate_detection_threshold),
matched_filter_lag_aggregator_(data_dumper_,
matched_filter_.GetMaxFilterLag(),
@ -71,7 +72,8 @@ absl::optional<DelayEstimate> EchoPathDelayEstimator::EstimateDelay(
data_dumper_->DumpWav("aec3_capture_decimator_output",
downsampled_capture.size(), downsampled_capture.data(),
16000 / down_sampling_factor_, 1);
matched_filter_.Update(render_buffer, downsampled_capture);
matched_filter_.Update(render_buffer, downsampled_capture,
matched_filter_lag_aggregator_.ReliableDelayFound());
absl::optional<DelayEstimate> aggregated_matched_filter_lag =
matched_filter_lag_aggregator_.Aggregate(

32
modules/audio_processing/aec3/matched_filter.cc
View File

@ -307,7 +307,8 @@ MatchedFilter::MatchedFilter(ApmDataDumper* data_dumper,
int num_matched_filters,
size_t alignment_shift_sub_blocks,
float excitation_limit,
float smoothing,
float smoothing_fast,
float smoothing_slow,
float matching_filter_threshold)
: data_dumper_(data_dumper),
optimization_(optimization),
@ -319,7 +320,8 @@ MatchedFilter::MatchedFilter(ApmDataDumper* data_dumper,
lag_estimates_(num_matched_filters),
filters_offsets_(num_matched_filters, 0),
excitation_limit_(excitation_limit),
smoothing_(smoothing),
smoothing_fast_(smoothing_fast),
smoothing_slow_(smoothing_slow),
matching_filter_threshold_(matching_filter_threshold) {
RTC_DCHECK(data_dumper);
RTC_DCHECK_LT(0, window_size_sub_blocks);
@ -340,10 +342,14 @@ void MatchedFilter::Reset() {
}
void MatchedFilter::Update(const DownsampledRenderBuffer& render_buffer,
rtc::ArrayView<const float> capture) {
rtc::ArrayView<const float> capture,
bool use_slow_smoothing) {
RTC_DCHECK_EQ(sub_block_size_, capture.size());
auto& y = capture;
const float smoothing =
use_slow_smoothing ? smoothing_slow_ : smoothing_fast_;
const float x2_sum_threshold =
filters_[0].size() * excitation_limit_ * excitation_limit_;
@ -360,25 +366,25 @@ void MatchedFilter::Update(const DownsampledRenderBuffer& render_buffer,
switch (optimization_) {
#if defined(WEBRTC_ARCH_X86_FAMILY)
case Aec3Optimization::kSse2:
aec3::MatchedFilterCore_SSE2(x_start_index, x2_sum_threshold,
smoothing_, render_buffer.buffer, y,
filters_[n], &filters_updated, &error_sum);
aec3::MatchedFilterCore_SSE2(x_start_index, x2_sum_threshold, smoothing,
render_buffer.buffer, y, filters_[n],
&filters_updated, &error_sum);
break;
case Aec3Optimization::kAvx2:
aec3::MatchedFilterCore_AVX2(x_start_index, x2_sum_threshold,
smoothing_, render_buffer.buffer, y,
filters_[n], &filters_updated, &error_sum);
aec3::MatchedFilterCore_AVX2(x_start_index, x2_sum_threshold, smoothing,
render_buffer.buffer, y, filters_[n],
&filters_updated, &error_sum);
break;
#endif
#if defined(WEBRTC_HAS_NEON)
case Aec3Optimization::kNeon:
aec3::MatchedFilterCore_NEON(x_start_index, x2_sum_threshold,
smoothing_, render_buffer.buffer, y,
filters_[n], &filters_updated, &error_sum);
aec3::MatchedFilterCore_NEON(x_start_index, x2_sum_threshold, smoothing,
render_buffer.buffer, y, filters_[n],
&filters_updated, &error_sum);
break;
#endif
default:
aec3::MatchedFilterCore(x_start_index, x2_sum_threshold, smoothing_,
aec3::MatchedFilterCore(x_start_index, x2_sum_threshold, smoothing,
render_buffer.buffer, y, filters_[n],
&filters_updated, &error_sum);
}

9
modules/audio_processing/aec3/matched_filter.h
View File

@ -100,7 +100,8 @@ class MatchedFilter {
int num_matched_filters,
size_t alignment_shift_sub_blocks,
float excitation_limit,
float smoothing,
float smoothing_fast,
float smoothing_slow,
float matching_filter_threshold);
MatchedFilter() = delete;
@ -111,7 +112,8 @@ class MatchedFilter {
// Updates the correlation with the values in the capture buffer.
void Update(const DownsampledRenderBuffer& render_buffer,
rtc::ArrayView<const float> capture);
rtc::ArrayView<const float> capture,
bool use_slow_smoothing);
// Resets the matched filter.
void Reset();
@ -140,7 +142,8 @@ class MatchedFilter {
std::vector<LagEstimate> lag_estimates_;
std::vector<size_t> filters_offsets_;
const float excitation_limit_;
const float smoothing_;
const float smoothing_fast_;
const float smoothing_slow_;
const float matching_filter_threshold_;
};

3
modules/audio_processing/aec3/matched_filter_lag_aggregator.h
View File

@ -45,6 +45,9 @@ class MatchedFilterLagAggregator {
absl::optional<DelayEstimate> Aggregate(
rtc::ArrayView<const MatchedFilter::LagEstimate> lag_estimates);
// Returns whether a reliable delay estimate has been found.
bool ReliableDelayFound() const { return significant_candidate_found_; }
private:
ApmDataDumper* const data_dumper_;
std::vector<int> histogram_;

15
modules/audio_processing/aec3/matched_filter_unittest.cc
View File

@ -206,6 +206,7 @@ TEST(MatchedFilter, LagEstimation) {
kWindowSizeSubBlocks, kNumMatchedFilters,
kAlignmentShiftSubBlocks, 150,
config.delay.delay_estimate_smoothing,
config.delay.delay_estimate_smoothing_delay_found,
config.delay.delay_candidate_detection_threshold);
std::unique_ptr<RenderDelayBuffer> render_delay_buffer(
@ -231,7 +232,7 @@ TEST(MatchedFilter, LagEstimation) {
downsampled_capture_data.data(), sub_block_size);
capture_decimator.Decimate(capture[0], downsampled_capture);
filter.Update(render_delay_buffer->GetDownsampledRenderBuffer(),
downsampled_capture);
downsampled_capture, false);
}
// Obtain the lag estimates.
@ -318,6 +319,7 @@ TEST(MatchedFilter, LagNotReliableForUncorrelatedRenderAndCapture) {
kWindowSizeSubBlocks, kNumMatchedFilters,
kAlignmentShiftSubBlocks, 150,
config.delay.delay_estimate_smoothing,
config.delay.delay_estimate_smoothing_delay_found,
config.delay.delay_candidate_detection_threshold);
// Analyze the correlation between render and capture.
@ -325,7 +327,8 @@ TEST(MatchedFilter, LagNotReliableForUncorrelatedRenderAndCapture) {
RandomizeSampleVector(&random_generator, render[0][0]);
RandomizeSampleVector(&random_generator, capture);
render_delay_buffer->Insert(render);
filter.Update(render_delay_buffer->GetDownsampledRenderBuffer(), capture);
filter.Update(render_delay_buffer->GetDownsampledRenderBuffer(), capture,
false);
}
// Obtain the lag estimates.
@ -361,6 +364,7 @@ TEST(MatchedFilter, LagNotUpdatedForLowLevelRender) {
kWindowSizeSubBlocks, kNumMatchedFilters,
kAlignmentShiftSubBlocks, 150,
config.delay.delay_estimate_smoothing,
config.delay.delay_estimate_smoothing_delay_found,
config.delay.delay_candidate_detection_threshold);
std::unique_ptr<RenderDelayBuffer> render_delay_buffer(
RenderDelayBuffer::Create(EchoCanceller3Config(), kSampleRateHz,
@ -379,7 +383,7 @@ TEST(MatchedFilter, LagNotUpdatedForLowLevelRender) {
sub_block_size);
capture_decimator.Decimate(capture[0], downsampled_capture);
filter.Update(render_delay_buffer->GetDownsampledRenderBuffer(),
downsampled_capture);
downsampled_capture, false);
}
// Obtain the lag estimates.
@ -407,6 +411,7 @@ TEST(MatchedFilter, NumberOfLagEstimates) {
MatchedFilter filter(&data_dumper, DetectOptimization(), sub_block_size,
32, num_matched_filters, 1, 150,
config.delay.delay_estimate_smoothing,
config.delay.delay_estimate_smoothing_delay_found,
config.delay.delay_candidate_detection_threshold);
EXPECT_EQ(num_matched_filters, filter.GetLagEstimates().size());
}
@ -421,6 +426,7 @@ TEST(MatchedFilterDeathTest, ZeroWindowSize) {
EchoCanceller3Config config;
EXPECT_DEATH(MatchedFilter(&data_dumper, DetectOptimization(), 16, 0, 1, 1,
150, config.delay.delay_estimate_smoothing,
config.delay.delay_estimate_smoothing_delay_found,
config.delay.delay_candidate_detection_threshold),
"");
}
@ -430,6 +436,7 @@ TEST(MatchedFilterDeathTest, NullDataDumper) {
EchoCanceller3Config config;
EXPECT_DEATH(MatchedFilter(nullptr, DetectOptimization(), 16, 1, 1, 1, 150,
config.delay.delay_estimate_smoothing,
config.delay.delay_estimate_smoothing_delay_found,
config.delay.delay_candidate_detection_threshold),
"");
}
@ -441,6 +448,7 @@ TEST(MatchedFilterDeathTest, DISABLED_BlockSizeMultipleOf4) {
EchoCanceller3Config config;
EXPECT_DEATH(MatchedFilter(&data_dumper, DetectOptimization(), 15, 1, 1, 1,
150, config.delay.delay_estimate_smoothing,
config.delay.delay_estimate_smoothing_delay_found,
config.delay.delay_candidate_detection_threshold),
"");
}
@ -453,6 +461,7 @@ TEST(MatchedFilterDeathTest, DISABLED_SubBlockSizeAddsUpToBlockSize) {
EchoCanceller3Config config;
EXPECT_DEATH(MatchedFilter(&data_dumper, DetectOptimization(), 12, 1, 1, 1,
150, config.delay.delay_estimate_smoothing,
config.delay.delay_estimate_smoothing_delay_found,
config.delay.delay_candidate_detection_threshold),
"");
}