1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299
|
/*
* 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/gain_controller2.h"
#include <algorithm>
#include <atomic>
#include <cmath>
#include <memory>
#include <optional>
#include "api/audio/audio_frame.h"
#include "api/audio/audio_processing.h"
#include "api/audio/audio_view.h"
#include "api/environment/environment.h"
#include "api/field_trials_view.h"
#include "common_audio/include/audio_util.h"
#include "modules/audio_processing/agc2/adaptive_digital_gain_controller.h"
#include "modules/audio_processing/agc2/agc2_common.h"
#include "modules/audio_processing/agc2/cpu_features.h"
#include "modules/audio_processing/agc2/input_volume_controller.h"
#include "modules/audio_processing/agc2/interpolated_gain_curve.h"
#include "modules/audio_processing/agc2/noise_level_estimator.h"
#include "modules/audio_processing/agc2/saturation_protector.h"
#include "modules/audio_processing/agc2/speech_level_estimator.h"
#include "modules/audio_processing/agc2/vad_wrapper.h"
#include "modules/audio_processing/audio_buffer.h"
#include "modules/audio_processing/logging/apm_data_dumper.h"
#include "rtc_base/checks.h"
#include "rtc_base/logging.h"
namespace webrtc {
namespace {
using Agc2Config = AudioProcessing::Config::GainController2;
using InputVolumeControllerConfig = InputVolumeController::Config;
constexpr int kLogLimiterStatsPeriodMs = 30'000;
constexpr int kFrameLengthMs = 10;
constexpr int kLogLimiterStatsPeriodNumFrames =
kLogLimiterStatsPeriodMs / kFrameLengthMs;
// Detects the available CPU features and applies any kill-switches.
AvailableCpuFeatures GetAllowedCpuFeatures(
const FieldTrialsView& field_trials) {
AvailableCpuFeatures features = GetAvailableCpuFeatures();
if (field_trials.IsEnabled("WebRTC-Agc2SimdSse2KillSwitch")) {
features.sse2 = false;
}
if (field_trials.IsEnabled("WebRTC-Agc2SimdAvx2KillSwitch")) {
features.avx2 = false;
}
if (field_trials.IsEnabled("WebRTC-Agc2SimdNeonKillSwitch")) {
features.neon = false;
}
return features;
}
// Peak and RMS audio levels in dBFS.
struct AudioLevels {
float peak_dbfs;
float rms_dbfs;
};
// Speech level info.
struct SpeechLevel {
bool is_confident;
float rms_dbfs;
};
// Computes the audio levels for the first channel in `frame`.
AudioLevels ComputeAudioLevels(DeinterleavedView<float> frame,
ApmDataDumper& data_dumper) {
float peak = 0.0f;
float rms = 0.0f;
for (const auto& x : frame[0]) {
peak = std::max(std::fabs(x), peak);
rms += x * x;
}
AudioLevels levels{
FloatS16ToDbfs(peak),
FloatS16ToDbfs(std::sqrt(rms / frame.samples_per_channel()))};
data_dumper.DumpRaw("agc2_input_rms_dbfs", levels.rms_dbfs);
data_dumper.DumpRaw("agc2_input_peak_dbfs", levels.peak_dbfs);
return levels;
}
} // namespace
std::atomic<int> GainController2::instance_count_(0);
GainController2::GainController2(
const Environment& env,
const Agc2Config& config,
const InputVolumeControllerConfig& input_volume_controller_config,
int sample_rate_hz,
int num_channels,
bool use_internal_vad)
: cpu_features_(GetAllowedCpuFeatures(env.field_trials())),
data_dumper_(instance_count_.fetch_add(1) + 1),
fixed_gain_applier_(
/*hard_clip_samples=*/false,
/*initial_gain_factor=*/DbToRatio(config.fixed_digital.gain_db)),
limiter_(&data_dumper_,
SampleRateToDefaultChannelSize(sample_rate_hz),
/*histogram_name_prefix=*/"Agc2"),
calls_since_last_limiter_log_(0) {
RTC_DCHECK(Validate(config));
data_dumper_.InitiateNewSetOfRecordings();
if (config.input_volume_controller.enabled ||
config.adaptive_digital.enabled) {
// Create dependencies.
speech_level_estimator_ = std::make_unique<SpeechLevelEstimator>(
&data_dumper_, config.adaptive_digital, kAdjacentSpeechFramesThreshold);
if (use_internal_vad)
vad_ = std::make_unique<VoiceActivityDetectorWrapper>(
kVadResetPeriodMs, cpu_features_, sample_rate_hz);
}
if (config.input_volume_controller.enabled) {
// Create controller.
input_volume_controller_ = std::make_unique<InputVolumeController>(
num_channels, input_volume_controller_config);
// TODO(bugs.webrtc.org/7494): Call `Initialize` in ctor and remove method.
input_volume_controller_->Initialize();
}
if (config.adaptive_digital.enabled) {
// Create dependencies.
noise_level_estimator_ = CreateNoiseFloorEstimator(&data_dumper_);
saturation_protector_ = CreateSaturationProtector(
kSaturationProtectorInitialHeadroomDb, kAdjacentSpeechFramesThreshold,
&data_dumper_);
// Create controller.
adaptive_digital_controller_ =
std::make_unique<AdaptiveDigitalGainController>(
&data_dumper_, config.adaptive_digital,
kAdjacentSpeechFramesThreshold);
}
}
GainController2::~GainController2() = default;
// TODO(webrtc:7494): Pass the flag also to the other components.
void GainController2::SetCaptureOutputUsed(bool capture_output_used) {
if (input_volume_controller_) {
input_volume_controller_->HandleCaptureOutputUsedChange(
capture_output_used);
}
}
void GainController2::SetFixedGainDb(float gain_db) {
const float gain_factor = DbToRatio(gain_db);
if (fixed_gain_applier_.GetGainFactor() != gain_factor) {
// Reset the limiter to quickly react on abrupt level changes caused by
// large changes of the fixed gain.
limiter_.Reset();
}
fixed_gain_applier_.SetGainFactor(gain_factor);
}
void GainController2::Analyze(int applied_input_volume,
const AudioBuffer& audio_buffer) {
recommended_input_volume_ = std::nullopt;
RTC_DCHECK_GE(applied_input_volume, 0);
RTC_DCHECK_LE(applied_input_volume, 255);
if (input_volume_controller_) {
input_volume_controller_->AnalyzeInputAudio(applied_input_volume,
audio_buffer);
}
}
void GainController2::Process(std::optional<float> speech_probability,
bool input_volume_changed,
AudioBuffer* audio) {
recommended_input_volume_ = std::nullopt;
data_dumper_.DumpRaw("agc2_applied_input_volume_changed",
input_volume_changed);
if (input_volume_changed) {
// Handle input volume changes.
if (speech_level_estimator_)
speech_level_estimator_->Reset();
if (saturation_protector_)
saturation_protector_->Reset();
}
DeinterleavedView<float> float_frame = audio->view();
// Compute speech probability.
if (vad_) {
// When the VAD component runs, `speech_probability` should not be specified
// because APM should not run the same VAD twice (as an APM sub-module and
// internally in AGC2).
RTC_DCHECK(!speech_probability.has_value());
speech_probability = vad_->Analyze(float_frame);
}
if (speech_probability.has_value()) {
RTC_DCHECK_GE(*speech_probability, 0.0f);
RTC_DCHECK_LE(*speech_probability, 1.0f);
}
// The speech probability may not be defined at this step (e.g., when the
// fixed digital controller alone is enabled).
if (speech_probability.has_value())
data_dumper_.DumpRaw("agc2_speech_probability", *speech_probability);
// Compute audio, noise and speech levels.
AudioLevels audio_levels = ComputeAudioLevels(float_frame, data_dumper_);
std::optional<float> noise_rms_dbfs;
if (noise_level_estimator_) {
// TODO(bugs.webrtc.org/7494): Pass `audio_levels` to remove duplicated
// computation in `noise_level_estimator_`.
noise_rms_dbfs = noise_level_estimator_->Analyze(float_frame);
}
std::optional<SpeechLevel> speech_level;
if (speech_level_estimator_) {
RTC_DCHECK(speech_probability.has_value());
speech_level_estimator_->Update(
audio_levels.rms_dbfs, audio_levels.peak_dbfs, *speech_probability);
speech_level =
SpeechLevel{.is_confident = speech_level_estimator_->is_confident(),
.rms_dbfs = speech_level_estimator_->level_dbfs()};
}
// Update the recommended input volume.
if (input_volume_controller_) {
RTC_DCHECK(speech_level.has_value());
RTC_DCHECK(speech_probability.has_value());
if (speech_probability.has_value()) {
recommended_input_volume_ =
input_volume_controller_->RecommendInputVolume(
*speech_probability,
speech_level->is_confident
? std::optional<float>(speech_level->rms_dbfs)
: std::nullopt);
}
}
if (adaptive_digital_controller_) {
RTC_DCHECK(saturation_protector_);
RTC_DCHECK(speech_probability.has_value());
RTC_DCHECK(speech_level.has_value());
saturation_protector_->Analyze(*speech_probability, audio_levels.peak_dbfs,
speech_level->rms_dbfs);
float headroom_db = saturation_protector_->HeadroomDb();
data_dumper_.DumpRaw("agc2_headroom_db", headroom_db);
float limiter_envelope_dbfs = FloatS16ToDbfs(limiter_.LastAudioLevel());
data_dumper_.DumpRaw("agc2_limiter_envelope_dbfs", limiter_envelope_dbfs);
RTC_DCHECK(noise_rms_dbfs.has_value());
adaptive_digital_controller_->Process(
/*info=*/{.speech_probability = *speech_probability,
.speech_level_dbfs = speech_level->rms_dbfs,
.speech_level_reliable = speech_level->is_confident,
.noise_rms_dbfs = *noise_rms_dbfs,
.headroom_db = headroom_db,
.limiter_envelope_dbfs = limiter_envelope_dbfs},
float_frame);
}
// TODO(bugs.webrtc.org/7494): Pass `audio_levels` to remove duplicated
// computation in `limiter_`.
fixed_gain_applier_.ApplyGain(float_frame);
limiter_.Process(float_frame);
// Periodically log limiter stats.
if (++calls_since_last_limiter_log_ == kLogLimiterStatsPeriodNumFrames) {
calls_since_last_limiter_log_ = 0;
InterpolatedGainCurve::Stats stats = limiter_.GetGainCurveStats();
RTC_LOG(LS_INFO) << "[AGC2] limiter stats"
<< " | identity: " << stats.look_ups_identity_region
<< " | knee: " << stats.look_ups_knee_region
<< " | limiter: " << stats.look_ups_limiter_region
<< " | saturation: " << stats.look_ups_saturation_region;
}
}
bool GainController2::Validate(
const AudioProcessing::Config::GainController2& config) {
const auto& fixed = config.fixed_digital;
const auto& adaptive = config.adaptive_digital;
return fixed.gain_db >= 0.0f && fixed.gain_db < 50.0f &&
adaptive.headroom_db >= 0.0f && adaptive.max_gain_db > 0.0f &&
adaptive.initial_gain_db >= 0.0f &&
adaptive.max_gain_change_db_per_second > 0.0f &&
adaptive.max_output_noise_level_dbfs <= 0.0f;
}
} // namespace webrtc
|