mirror of
https://github.com/mollyim/webrtc.git
synced 2025-05-17 23:57:59 +01:00

This CL separates the NLP gain computation for the different variants of echo estimation. This simplifies the setting of tuning parameters, with resulting transparency improvements and increased echo removal performance. Bug: webrtc:8359 Change-Id: I9b97064396fb6f6e2f418ce534573f68694390a1 Reviewed-on: https://webrtc-review.googlesource.com/7613 Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org> Commit-Queue: Per Åhgren <peah@webrtc.org> Cr-Commit-Position: refs/heads/master@{#20209}
262 lines
9.4 KiB
C++
262 lines
9.4 KiB
C++
/*
|
|
* Copyright (c) 2017 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 "modules/audio_processing/aec3/residual_echo_estimator.h"
|
|
|
|
#include <numeric>
|
|
#include <vector>
|
|
|
|
#include "rtc_base/checks.h"
|
|
|
|
namespace webrtc {
|
|
namespace {
|
|
|
|
// Estimates the echo generating signal power as gated maximal power over a time
|
|
// window.
|
|
void EchoGeneratingPower(const RenderBuffer& render_buffer,
|
|
size_t min_delay,
|
|
size_t max_delay,
|
|
std::array<float, kFftLengthBy2Plus1>* X2) {
|
|
X2->fill(0.f);
|
|
for (size_t k = min_delay; k <= max_delay; ++k) {
|
|
std::transform(X2->begin(), X2->end(), render_buffer.Spectrum(k).begin(),
|
|
X2->begin(),
|
|
[](float a, float b) { return std::max(a, b); });
|
|
}
|
|
|
|
// Apply soft noise gate of -78 dBFS.
|
|
static constexpr float kNoiseGatePower = 27509.42f;
|
|
std::for_each(X2->begin(), X2->end(), [](float& a) {
|
|
if (kNoiseGatePower > a) {
|
|
a = std::max(0.f, a - 0.3f * (kNoiseGatePower - a));
|
|
}
|
|
});
|
|
}
|
|
|
|
constexpr int kNoiseFloorCounterMax = 50;
|
|
constexpr float kNoiseFloorMin = 10.f * 10.f * 128.f * 128.f;
|
|
|
|
// Updates estimate for the power of the stationary noise component in the
|
|
// render signal.
|
|
void RenderNoisePower(
|
|
const RenderBuffer& render_buffer,
|
|
std::array<float, kFftLengthBy2Plus1>* X2_noise_floor,
|
|
std::array<int, kFftLengthBy2Plus1>* X2_noise_floor_counter) {
|
|
RTC_DCHECK(X2_noise_floor);
|
|
RTC_DCHECK(X2_noise_floor_counter);
|
|
|
|
const auto render_power = render_buffer.Spectrum(0);
|
|
RTC_DCHECK_EQ(X2_noise_floor->size(), render_power.size());
|
|
RTC_DCHECK_EQ(X2_noise_floor_counter->size(), render_power.size());
|
|
|
|
// Estimate the stationary noise power in a minimum statistics manner.
|
|
for (size_t k = 0; k < render_power.size(); ++k) {
|
|
// Decrease rapidly.
|
|
if (render_power[k] < (*X2_noise_floor)[k]) {
|
|
(*X2_noise_floor)[k] = render_power[k];
|
|
(*X2_noise_floor_counter)[k] = 0;
|
|
} else {
|
|
// Increase in a delayed, leaky manner.
|
|
if ((*X2_noise_floor_counter)[k] >= kNoiseFloorCounterMax) {
|
|
(*X2_noise_floor)[k] =
|
|
std::max((*X2_noise_floor)[k] * 1.1f, kNoiseFloorMin);
|
|
} else {
|
|
++(*X2_noise_floor_counter)[k];
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
// Assume a minimum echo path gain of -33 dB for headsets.
|
|
constexpr float kHeadsetEchoPathGain = 0.0005f;
|
|
|
|
} // namespace
|
|
|
|
ResidualEchoEstimator::ResidualEchoEstimator(
|
|
const AudioProcessing::Config::EchoCanceller3& config)
|
|
: config_(config) {
|
|
Reset();
|
|
}
|
|
|
|
ResidualEchoEstimator::~ResidualEchoEstimator() = default;
|
|
|
|
void ResidualEchoEstimator::Estimate(
|
|
const AecState& aec_state,
|
|
const RenderBuffer& render_buffer,
|
|
const std::array<float, kFftLengthBy2Plus1>& S2_linear,
|
|
const std::array<float, kFftLengthBy2Plus1>& Y2,
|
|
std::array<float, kFftLengthBy2Plus1>* R2) {
|
|
RTC_DCHECK(R2);
|
|
|
|
const rtc::Optional<size_t> delay =
|
|
aec_state.ExternalDelay()
|
|
? (aec_state.FilterDelay() ? aec_state.FilterDelay()
|
|
: aec_state.ExternalDelay())
|
|
: rtc::Optional<size_t>();
|
|
|
|
// Estimate the power of the stationary noise in the render signal.
|
|
RenderNoisePower(render_buffer, &X2_noise_floor_, &X2_noise_floor_counter_);
|
|
|
|
// Estimate the residual echo power.
|
|
|
|
if (aec_state.LinearEchoEstimate()) {
|
|
RTC_DCHECK(aec_state.FilterDelay());
|
|
const int filter_delay = *aec_state.FilterDelay();
|
|
LinearEstimate(S2_linear, aec_state.Erle(), filter_delay, R2);
|
|
AddEchoReverb(S2_linear, aec_state.SaturatedEcho(), filter_delay,
|
|
aec_state.ReverbDecay(), R2);
|
|
} else {
|
|
// Estimate the echo generating signal power.
|
|
std::array<float, kFftLengthBy2Plus1> X2;
|
|
if (aec_state.ExternalDelay() && aec_state.FilterDelay()) {
|
|
RTC_DCHECK(delay);
|
|
const int delay_use = static_cast<int>(*delay);
|
|
|
|
// Computes the spectral power over the blocks surrounding the delay.
|
|
RTC_DCHECK_LT(delay_use, kResidualEchoPowerRenderWindowSize);
|
|
EchoGeneratingPower(
|
|
render_buffer, std::max(0, delay_use - 1),
|
|
std::min(kResidualEchoPowerRenderWindowSize - 1, delay_use + 1), &X2);
|
|
} else {
|
|
// Computes the spectral power over the latest blocks.
|
|
EchoGeneratingPower(render_buffer, 0,
|
|
kResidualEchoPowerRenderWindowSize - 1, &X2);
|
|
}
|
|
|
|
// Subtract the stationary noise power to avoid stationary noise causing
|
|
// excessive echo suppression.
|
|
std::transform(
|
|
X2.begin(), X2.end(), X2_noise_floor_.begin(), X2.begin(),
|
|
[](float a, float b) { return std::max(0.f, a - 10.f * b); });
|
|
|
|
NonLinearEstimate(aec_state.HeadsetDetected(), X2, Y2, R2);
|
|
AddEchoReverb(*R2, aec_state.SaturatedEcho(),
|
|
std::min(static_cast<size_t>(kAdaptiveFilterLength),
|
|
delay.value_or(kAdaptiveFilterLength)),
|
|
aec_state.ReverbDecay(), R2);
|
|
}
|
|
|
|
// If the echo is deemed inaudible, set the residual echo to zero.
|
|
if (aec_state.InaudibleEcho() &&
|
|
(aec_state.ExternalDelay() || aec_state.HeadsetDetected())) {
|
|
R2->fill(0.f);
|
|
}
|
|
|
|
// If the echo is saturated, estimate the echo power as the maximum echo power
|
|
// with a leakage factor.
|
|
if (aec_state.SaturatedEcho()) {
|
|
R2->fill((*std::max_element(R2->begin(), R2->end())) * 100.f);
|
|
}
|
|
|
|
std::copy(R2->begin(), R2->end(), R2_old_.begin());
|
|
}
|
|
|
|
void ResidualEchoEstimator::Reset() {
|
|
X2_noise_floor_counter_.fill(kNoiseFloorCounterMax);
|
|
X2_noise_floor_.fill(kNoiseFloorMin);
|
|
R2_reverb_.fill(0.f);
|
|
R2_old_.fill(0.f);
|
|
R2_hold_counter_.fill(0.f);
|
|
for (auto& S2_k : S2_old_) {
|
|
S2_k.fill(0.f);
|
|
}
|
|
}
|
|
|
|
void ResidualEchoEstimator::LinearEstimate(
|
|
const std::array<float, kFftLengthBy2Plus1>& S2_linear,
|
|
const std::array<float, kFftLengthBy2Plus1>& erle,
|
|
size_t delay,
|
|
std::array<float, kFftLengthBy2Plus1>* R2) {
|
|
std::fill(R2_hold_counter_.begin(), R2_hold_counter_.end(), 10.f);
|
|
std::transform(erle.begin(), erle.end(), S2_linear.begin(), R2->begin(),
|
|
[](float a, float b) {
|
|
RTC_DCHECK_LT(0.f, a);
|
|
return b / a;
|
|
});
|
|
}
|
|
|
|
void ResidualEchoEstimator::NonLinearEstimate(
|
|
bool headset_detected,
|
|
const std::array<float, kFftLengthBy2Plus1>& X2,
|
|
const std::array<float, kFftLengthBy2Plus1>& Y2,
|
|
std::array<float, kFftLengthBy2Plus1>* R2) {
|
|
// Choose gains.
|
|
const float echo_path_gain_lf =
|
|
headset_detected ? kHeadsetEchoPathGain : config_.param.ep_strength.lf;
|
|
const float echo_path_gain_mf =
|
|
headset_detected ? kHeadsetEchoPathGain : config_.param.ep_strength.mf;
|
|
const float echo_path_gain_hf =
|
|
headset_detected ? kHeadsetEchoPathGain : config_.param.ep_strength.hf;
|
|
|
|
// Compute preliminary residual echo.
|
|
std::transform(
|
|
X2.begin(), X2.begin() + 12, R2->begin(),
|
|
[echo_path_gain_lf](float a) { return a * echo_path_gain_lf; });
|
|
std::transform(
|
|
X2.begin() + 12, X2.begin() + 25, R2->begin() + 12,
|
|
[echo_path_gain_mf](float a) { return a * echo_path_gain_mf; });
|
|
std::transform(
|
|
X2.begin() + 25, X2.end(), R2->begin() + 25,
|
|
[echo_path_gain_hf](float a) { return a * echo_path_gain_hf; });
|
|
|
|
for (size_t k = 0; k < R2->size(); ++k) {
|
|
// Update hold counter.
|
|
R2_hold_counter_[k] = R2_old_[k] < (*R2)[k] ? 0 : R2_hold_counter_[k] + 1;
|
|
|
|
// Compute the residual echo by holding a maximum echo powers and an echo
|
|
// fading corresponding to a room with an RT60 value of about 50 ms.
|
|
(*R2)[k] = R2_hold_counter_[k] < 2
|
|
? std::max((*R2)[k], R2_old_[k])
|
|
: std::min((*R2)[k] + R2_old_[k] * 0.1f, Y2[k]);
|
|
}
|
|
}
|
|
|
|
void ResidualEchoEstimator::AddEchoReverb(
|
|
const std::array<float, kFftLengthBy2Plus1>& S2,
|
|
bool saturated_echo,
|
|
size_t delay,
|
|
float reverb_decay_factor,
|
|
std::array<float, kFftLengthBy2Plus1>* R2) {
|
|
// Compute the decay factor for how much the echo has decayed before leaving
|
|
// the region covered by the linear model.
|
|
auto integer_power = [](float base, int exp) {
|
|
float result = 1.f;
|
|
for (int k = 0; k < exp; ++k) {
|
|
result *= base;
|
|
}
|
|
return result;
|
|
};
|
|
RTC_DCHECK_LE(delay, S2_old_.size());
|
|
const float reverb_decay_for_delay =
|
|
integer_power(reverb_decay_factor, S2_old_.size() - delay);
|
|
|
|
// Update the estimate of the reverberant residual echo power.
|
|
S2_old_index_ = S2_old_index_ > 0 ? S2_old_index_ - 1 : S2_old_.size() - 1;
|
|
const auto& S2_end = S2_old_[S2_old_index_];
|
|
std::transform(
|
|
S2_end.begin(), S2_end.end(), R2_reverb_.begin(), R2_reverb_.begin(),
|
|
[reverb_decay_for_delay, reverb_decay_factor](float a, float b) {
|
|
return (b + a * reverb_decay_for_delay) * reverb_decay_factor;
|
|
});
|
|
|
|
// Update the buffer of old echo powers.
|
|
if (saturated_echo) {
|
|
S2_old_[S2_old_index_].fill((*std::max_element(S2.begin(), S2.end())) *
|
|
100.f);
|
|
} else {
|
|
std::copy(S2.begin(), S2.end(), S2_old_[S2_old_index_].begin());
|
|
}
|
|
|
|
// Add the power of the echo reverb to the residual echo power.
|
|
std::transform(R2->begin(), R2->end(), R2_reverb_.begin(), R2->begin(),
|
|
std::plus<float>());
|
|
}
|
|
|
|
} // namespace webrtc
|