Tommi ff2bf4b195 Update FrameCombiner to use audio view methods for interleaved buffers
Along the way slightly simplify the class interface since views
carry audio properties. Also, now allocating FrameCombiner allocates
the mixing buffer in the same allocation.

Bug: chromium:335805780
Change-Id: Id7a76b040c11064e1e4daf01a371328769162554
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/352502
Commit-Queue: Tomas Gunnarsson <tommi@webrtc.org>
Reviewed-by: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#42465}
2024-06-12 09:44:40 +00:00

157 lines
5.6 KiB
C++

/*
* Copyright (c) 2018 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/agc2/limiter.h"
#include <algorithm>
#include <array>
#include <cmath>
#include "absl/strings/string_view.h"
#include "api/array_view.h"
#include "modules/audio_processing/agc2/agc2_common.h"
#include "modules/audio_processing/logging/apm_data_dumper.h"
#include "rtc_base/checks.h"
#include "rtc_base/numerics/safe_conversions.h"
#include "rtc_base/numerics/safe_minmax.h"
namespace webrtc {
namespace {
// This constant affects the way scaling factors are interpolated for the first
// sub-frame of a frame. Only in the case in which the first sub-frame has an
// estimated level which is greater than the that of the previous analyzed
// sub-frame, linear interpolation is replaced with a power function which
// reduces the chances of over-shooting (and hence saturation), however reducing
// the fixed gain effectiveness.
constexpr float kAttackFirstSubframeInterpolationPower = 8.0f;
void InterpolateFirstSubframe(float last_factor,
float current_factor,
rtc::ArrayView<float> subframe) {
const int n = rtc::dchecked_cast<int>(subframe.size());
constexpr float p = kAttackFirstSubframeInterpolationPower;
for (int i = 0; i < n; ++i) {
subframe[i] = std::pow(1.f - i / n, p) * (last_factor - current_factor) +
current_factor;
}
}
void ComputePerSampleSubframeFactors(
const std::array<float, kSubFramesInFrame + 1>& scaling_factors,
int samples_per_channel,
rtc::ArrayView<float> per_sample_scaling_factors) {
const int num_subframes = scaling_factors.size() - 1;
const int subframe_size =
rtc::CheckedDivExact(samples_per_channel, num_subframes);
// Handle first sub-frame differently in case of attack.
const bool is_attack = scaling_factors[0] > scaling_factors[1];
if (is_attack) {
InterpolateFirstSubframe(
scaling_factors[0], scaling_factors[1],
rtc::ArrayView<float>(
per_sample_scaling_factors.subview(0, subframe_size)));
}
for (int i = is_attack ? 1 : 0; i < num_subframes; ++i) {
const int subframe_start = i * subframe_size;
const float scaling_start = scaling_factors[i];
const float scaling_end = scaling_factors[i + 1];
const float scaling_diff = (scaling_end - scaling_start) / subframe_size;
for (int j = 0; j < subframe_size; ++j) {
per_sample_scaling_factors[subframe_start + j] =
scaling_start + scaling_diff * j;
}
}
}
void ScaleSamples(MonoView<const float> per_sample_scaling_factors,
AudioFrameView<float> signal) {
const int samples_per_channel = signal.samples_per_channel();
RTC_DCHECK_EQ(samples_per_channel,
SamplesPerChannel(per_sample_scaling_factors));
for (int i = 0; i < signal.num_channels(); ++i) {
MonoView<float> channel = signal.channel(i);
for (int j = 0; j < samples_per_channel; ++j) {
channel[j] = rtc::SafeClamp(channel[j] * per_sample_scaling_factors[j],
kMinFloatS16Value, kMaxFloatS16Value);
}
}
}
void CheckLimiterSampleRate(int sample_rate_hz) {
// Check that per_sample_scaling_factors_ is large enough.
RTC_DCHECK_LE(sample_rate_hz,
kMaximalNumberOfSamplesPerChannel * 1000 / kFrameDurationMs);
}
} // namespace
Limiter::Limiter(int sample_rate_hz,
ApmDataDumper* apm_data_dumper,
absl::string_view histogram_name)
: interp_gain_curve_(apm_data_dumper, histogram_name),
level_estimator_(sample_rate_hz, apm_data_dumper),
apm_data_dumper_(apm_data_dumper) {
CheckLimiterSampleRate(sample_rate_hz);
}
Limiter::~Limiter() = default;
void Limiter::Process(AudioFrameView<float> signal) {
const std::array<float, kSubFramesInFrame> level_estimate =
level_estimator_.ComputeLevel(signal);
RTC_DCHECK_EQ(level_estimate.size() + 1, scaling_factors_.size());
scaling_factors_[0] = last_scaling_factor_;
std::transform(level_estimate.begin(), level_estimate.end(),
scaling_factors_.begin() + 1, [this](float x) {
return interp_gain_curve_.LookUpGainToApply(x);
});
const int samples_per_channel = signal.samples_per_channel();
RTC_DCHECK_LE(samples_per_channel, kMaximalNumberOfSamplesPerChannel);
auto per_sample_scaling_factors =
MonoView<float>(&per_sample_scaling_factors_[0], samples_per_channel);
ComputePerSampleSubframeFactors(scaling_factors_, samples_per_channel,
per_sample_scaling_factors);
ScaleSamples(per_sample_scaling_factors, signal);
last_scaling_factor_ = scaling_factors_.back();
// Dump data for debug.
apm_data_dumper_->DumpRaw("agc2_limiter_last_scaling_factor",
last_scaling_factor_);
apm_data_dumper_->DumpRaw(
"agc2_limiter_region",
static_cast<int>(interp_gain_curve_.get_stats().region));
}
InterpolatedGainCurve::Stats Limiter::GetGainCurveStats() const {
return interp_gain_curve_.get_stats();
}
void Limiter::SetSampleRate(int sample_rate_hz) {
CheckLimiterSampleRate(sample_rate_hz);
level_estimator_.SetSampleRate(sample_rate_hz);
}
void Limiter::Reset() {
level_estimator_.Reset();
}
float Limiter::LastAudioLevel() const {
return level_estimator_.LastAudioLevel();
}
} // namespace webrtc