Repo created

TMessagesProj/jni/voip/webrtc/audio/utility/audio_frame_operations.cc

@@ -0,0 +1,294 @@
+/*
+ *  Copyright (c) 2012 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 "audio/utility/audio_frame_operations.h"
+
+#include <string.h>
+
+#include <algorithm>
+#include <cstdint>
+#include <utility>
+
+#include "common_audio/include/audio_util.h"
+#include "rtc_base/checks.h"
+#include "rtc_base/numerics/safe_conversions.h"
+
+namespace webrtc {
+namespace {
+
+// 2.7ms @ 48kHz, 4ms @ 32kHz, 8ms @ 16kHz.
+const size_t kMuteFadeFrames = 128;
+const float kMuteFadeInc = 1.0f / kMuteFadeFrames;
+
+}  // namespace
+
+void AudioFrameOperations::Add(const AudioFrame& frame_to_add,
+                               AudioFrame* result_frame) {
+  // Sanity check.
+  RTC_DCHECK(result_frame);
+  RTC_DCHECK_GT(result_frame->num_channels_, 0);
+  RTC_DCHECK_EQ(result_frame->num_channels_, frame_to_add.num_channels_);
+
+  bool no_previous_data = result_frame->muted();
+  if (result_frame->samples_per_channel_ != frame_to_add.samples_per_channel_) {
+    // Special case we have no data to start with.
+    RTC_DCHECK_EQ(result_frame->samples_per_channel_, 0);
+    result_frame->samples_per_channel_ = frame_to_add.samples_per_channel_;
+    no_previous_data = true;
+  }
+
+  if (result_frame->vad_activity_ == AudioFrame::kVadActive ||
+      frame_to_add.vad_activity_ == AudioFrame::kVadActive) {
+    result_frame->vad_activity_ = AudioFrame::kVadActive;
+  } else if (result_frame->vad_activity_ == AudioFrame::kVadUnknown ||
+             frame_to_add.vad_activity_ == AudioFrame::kVadUnknown) {
+    result_frame->vad_activity_ = AudioFrame::kVadUnknown;
+  }
+
+  if (result_frame->speech_type_ != frame_to_add.speech_type_)
+    result_frame->speech_type_ = AudioFrame::kUndefined;
+
+  if (!frame_to_add.muted()) {
+    const int16_t* in_data = frame_to_add.data();
+    int16_t* out_data = result_frame->mutable_data();
+    size_t length =
+        frame_to_add.samples_per_channel_ * frame_to_add.num_channels_;
+    if (no_previous_data) {
+      std::copy(in_data, in_data + length, out_data);
+    } else {
+      for (size_t i = 0; i < length; i++) {
+        const int32_t wrap_guard = static_cast<int32_t>(out_data[i]) +
+                                   static_cast<int32_t>(in_data[i]);
+        out_data[i] = rtc::saturated_cast<int16_t>(wrap_guard);
+      }
+    }
+  }
+}
+
+int AudioFrameOperations::MonoToStereo(AudioFrame* frame) {
+  if (frame->num_channels_ != 1) {
+    return -1;
+  }
+  UpmixChannels(2, frame);
+  return 0;
+}
+
+int AudioFrameOperations::StereoToMono(AudioFrame* frame) {
+  if (frame->num_channels_ != 2) {
+    return -1;
+  }
+  DownmixChannels(1, frame);
+  return frame->num_channels_ == 1 ? 0 : -1;
+}
+
+void AudioFrameOperations::QuadToStereo(const int16_t* src_audio,
+                                        size_t samples_per_channel,
+                                        int16_t* dst_audio) {
+  for (size_t i = 0; i < samples_per_channel; i++) {
+    dst_audio[i * 2] =
+        (static_cast<int32_t>(src_audio[4 * i]) + src_audio[4 * i + 1]) >> 1;
+    dst_audio[i * 2 + 1] =
+        (static_cast<int32_t>(src_audio[4 * i + 2]) + src_audio[4 * i + 3]) >>
+        1;
+  }
+}
+
+int AudioFrameOperations::QuadToStereo(AudioFrame* frame) {
+  if (frame->num_channels_ != 4) {
+    return -1;
+  }
+
+  RTC_DCHECK_LE(frame->samples_per_channel_ * 4,
+                AudioFrame::kMaxDataSizeSamples);
+
+  if (!frame->muted()) {
+    QuadToStereo(frame->data(), frame->samples_per_channel_,
+                 frame->mutable_data());
+  }
+  frame->num_channels_ = 2;
+
+  return 0;
+}
+
+void AudioFrameOperations::DownmixChannels(const int16_t* src_audio,
+                                           size_t src_channels,
+                                           size_t samples_per_channel,
+                                           size_t dst_channels,
+                                           int16_t* dst_audio) {
+  if (src_channels > 1 && dst_channels == 1) {
+    DownmixInterleavedToMono(src_audio, samples_per_channel, src_channels,
+                             dst_audio);
+    return;
+  } else if (src_channels == 4 && dst_channels == 2) {
+    QuadToStereo(src_audio, samples_per_channel, dst_audio);
+    return;
+  }
+
+  RTC_DCHECK_NOTREACHED() << "src_channels: " << src_channels
+                          << ", dst_channels: " << dst_channels;
+}
+
+void AudioFrameOperations::DownmixChannels(size_t dst_channels,
+                                           AudioFrame* frame) {
+  RTC_DCHECK_LE(frame->samples_per_channel_ * frame->num_channels_,
+                AudioFrame::kMaxDataSizeSamples);
+  if (frame->num_channels_ > 1 && dst_channels == 1) {
+    if (!frame->muted()) {
+      DownmixInterleavedToMono(frame->data(), frame->samples_per_channel_,
+                               frame->num_channels_, frame->mutable_data());
+    }
+    frame->num_channels_ = 1;
+  } else if (frame->num_channels_ == 4 && dst_channels == 2) {
+    int err = QuadToStereo(frame);
+    RTC_DCHECK_EQ(err, 0);
+  } else {
+    RTC_DCHECK_NOTREACHED() << "src_channels: " << frame->num_channels_
+                            << ", dst_channels: " << dst_channels;
+  }
+}
+
+void AudioFrameOperations::UpmixChannels(size_t target_number_of_channels,
+                                         AudioFrame* frame) {
+  RTC_DCHECK_EQ(frame->num_channels_, 1);
+  RTC_DCHECK_LE(frame->samples_per_channel_ * target_number_of_channels,
+                AudioFrame::kMaxDataSizeSamples);
+
+  if (frame->num_channels_ != 1 ||
+      frame->samples_per_channel_ * target_number_of_channels >
+          AudioFrame::kMaxDataSizeSamples) {
+    return;
+  }
+
+  if (!frame->muted()) {
+    // Up-mixing done in place. Going backwards through the frame ensure nothing
+    // is irrevocably overwritten.
+    int16_t* frame_data = frame->mutable_data();
+    for (int i = frame->samples_per_channel_ - 1; i >= 0; i--) {
+      for (size_t j = 0; j < target_number_of_channels; ++j) {
+        frame_data[target_number_of_channels * i + j] = frame_data[i];
+      }
+    }
+  }
+  frame->num_channels_ = target_number_of_channels;
+}
+
+void AudioFrameOperations::SwapStereoChannels(AudioFrame* frame) {
+  RTC_DCHECK(frame);
+  if (frame->num_channels_ != 2 || frame->muted()) {
+    return;
+  }
+
+  int16_t* frame_data = frame->mutable_data();
+  for (size_t i = 0; i < frame->samples_per_channel_ * 2; i += 2) {
+    std::swap(frame_data[i], frame_data[i + 1]);
+  }
+}
+
+void AudioFrameOperations::Mute(AudioFrame* frame,
+                                bool previous_frame_muted,
+                                bool current_frame_muted) {
+  RTC_DCHECK(frame);
+  if (!previous_frame_muted && !current_frame_muted) {
+    // Not muted, don't touch.
+  } else if (previous_frame_muted && current_frame_muted) {
+    // Frame fully muted.
+    size_t total_samples = frame->samples_per_channel_ * frame->num_channels_;
+    RTC_DCHECK_GE(AudioFrame::kMaxDataSizeSamples, total_samples);
+    frame->Mute();
+  } else {
+    // Fade is a no-op on a muted frame.
+    if (frame->muted()) {
+      return;
+    }
+
+    // Limit number of samples to fade, if frame isn't long enough.
+    size_t count = kMuteFadeFrames;
+    float inc = kMuteFadeInc;
+    if (frame->samples_per_channel_ < kMuteFadeFrames) {
+      count = frame->samples_per_channel_;
+      if (count > 0) {
+        inc = 1.0f / count;
+      }
+    }
+
+    size_t start = 0;
+    size_t end = count;
+    float start_g = 0.0f;
+    if (current_frame_muted) {
+      // Fade out the last `count` samples of frame.
+      RTC_DCHECK(!previous_frame_muted);
+      start = frame->samples_per_channel_ - count;
+      end = frame->samples_per_channel_;
+      start_g = 1.0f;
+      inc = -inc;
+    } else {
+      // Fade in the first `count` samples of frame.
+      RTC_DCHECK(previous_frame_muted);
+    }
+
+    // Perform fade.
+    int16_t* frame_data = frame->mutable_data();
+    size_t channels = frame->num_channels_;
+    for (size_t j = 0; j < channels; ++j) {
+      float g = start_g;
+      for (size_t i = start * channels; i < end * channels; i += channels) {
+        g += inc;
+        frame_data[i + j] *= g;
+      }
+    }
+  }
+}
+
+void AudioFrameOperations::Mute(AudioFrame* frame) {
+  Mute(frame, true, true);
+}
+
+void AudioFrameOperations::ApplyHalfGain(AudioFrame* frame) {
+  RTC_DCHECK(frame);
+  RTC_DCHECK_GT(frame->num_channels_, 0);
+  if (frame->num_channels_ < 1 || frame->muted()) {
+    return;
+  }
+
+  int16_t* frame_data = frame->mutable_data();
+  for (size_t i = 0; i < frame->samples_per_channel_ * frame->num_channels_;
+       i++) {
+    frame_data[i] = frame_data[i] >> 1;
+  }
+}
+
+int AudioFrameOperations::Scale(float left, float right, AudioFrame* frame) {
+  if (frame->num_channels_ != 2) {
+    return -1;
+  } else if (frame->muted()) {
+    return 0;
+  }
+
+  int16_t* frame_data = frame->mutable_data();
+  for (size_t i = 0; i < frame->samples_per_channel_; i++) {
+    frame_data[2 * i] = static_cast<int16_t>(left * frame_data[2 * i]);
+    frame_data[2 * i + 1] = static_cast<int16_t>(right * frame_data[2 * i + 1]);
+  }
+  return 0;
+}
+
+int AudioFrameOperations::ScaleWithSat(float scale, AudioFrame* frame) {
+  if (frame->muted()) {
+    return 0;
+  }
+
+  int16_t* frame_data = frame->mutable_data();
+  for (size_t i = 0; i < frame->samples_per_channel_ * frame->num_channels_;
+       i++) {
+    frame_data[i] = rtc::saturated_cast<int16_t>(scale * frame_data[i]);
+  }
+  return 0;
+}
+}  // namespace webrtc

TMessagesProj/jni/voip/webrtc/audio/utility/audio_frame_operations.h

@@ -0,0 +1,107 @@
+/*
+ *  Copyright (c) 2012 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.
+ */
+
+#ifndef AUDIO_UTILITY_AUDIO_FRAME_OPERATIONS_H_
+#define AUDIO_UTILITY_AUDIO_FRAME_OPERATIONS_H_
+
+#include <stddef.h>
+#include <stdint.h>
+
+#include "absl/base/attributes.h"
+#include "api/audio/audio_frame.h"
+
+namespace webrtc {
+
+// TODO(andrew): consolidate this with utility.h and audio_frame_manipulator.h.
+// Change reference parameters to pointers. Consider using a namespace rather
+// than a class.
+class AudioFrameOperations {
+ public:
+  // Add samples in `frame_to_add` with samples in `result_frame`
+  // putting the results in `results_frame`.  The fields
+  // `vad_activity_` and `speech_type_` of the result frame are
+  // updated. If `result_frame` is empty (`samples_per_channel_`==0),
+  // the samples in `frame_to_add` are added to it.  The number of
+  // channels and number of samples per channel must match except when
+  // `result_frame` is empty.
+  static void Add(const AudioFrame& frame_to_add, AudioFrame* result_frame);
+
+  // `frame.num_channels_` will be updated. This version checks for sufficient
+  // buffer size and that `num_channels_` is mono. Use UpmixChannels
+  // instead. TODO(bugs.webrtc.org/8649): remove.
+  ABSL_DEPRECATED("bugs.webrtc.org/8649")
+  static int MonoToStereo(AudioFrame* frame);
+
+  // `frame.num_channels_` will be updated. This version checks that
+  // `num_channels_` is stereo. Use DownmixChannels
+  // instead. TODO(bugs.webrtc.org/8649): remove.
+  ABSL_DEPRECATED("bugs.webrtc.org/8649")
+  static int StereoToMono(AudioFrame* frame);
+
+  // Downmixes 4 channels `src_audio` to stereo `dst_audio`. This is an in-place
+  // operation, meaning `src_audio` and `dst_audio` may point to the same
+  // buffer.
+  static void QuadToStereo(const int16_t* src_audio,
+                           size_t samples_per_channel,
+                           int16_t* dst_audio);
+
+  // `frame.num_channels_` will be updated. This version checks that
+  // `num_channels_` is 4 channels.
+  static int QuadToStereo(AudioFrame* frame);
+
+  // Downmixes `src_channels` `src_audio` to `dst_channels` `dst_audio`.
+  // This is an in-place operation, meaning `src_audio` and `dst_audio`
+  // may point to the same buffer. Supported channel combinations are
+  // Stereo to Mono, Quad to Mono, and Quad to Stereo.
+  static void DownmixChannels(const int16_t* src_audio,
+                              size_t src_channels,
+                              size_t samples_per_channel,
+                              size_t dst_channels,
+                              int16_t* dst_audio);
+
+  // `frame.num_channels_` will be updated. This version checks that
+  // `num_channels_` and `dst_channels` are valid and performs relevant downmix.
+  // Supported channel combinations are N channels to Mono, and Quad to Stereo.
+  static void DownmixChannels(size_t dst_channels, AudioFrame* frame);
+
+  // `frame.num_channels_` will be updated. This version checks that
+  // `num_channels_` and `dst_channels` are valid and performs relevant
+  // downmix. Supported channel combinations are Mono to N
+  // channels. The single channel is replicated.
+  static void UpmixChannels(size_t target_number_of_channels,
+                            AudioFrame* frame);
+
+  // Swap the left and right channels of `frame`. Fails silently if `frame` is
+  // not stereo.
+  static void SwapStereoChannels(AudioFrame* frame);
+
+  // Conditionally zero out contents of `frame` for implementing audio mute:
+  //  `previous_frame_muted` &&  `current_frame_muted` - Zero out whole frame.
+  //  `previous_frame_muted` && !`current_frame_muted` - Fade-in at frame start.
+  // !`previous_frame_muted` &&  `current_frame_muted` - Fade-out at frame end.
+  // !`previous_frame_muted` && !`current_frame_muted` - Leave frame untouched.
+  static void Mute(AudioFrame* frame,
+                   bool previous_frame_muted,
+                   bool current_frame_muted);
+
+  // Zero out contents of frame.
+  static void Mute(AudioFrame* frame);
+
+  // Halve samples in `frame`.
+  static void ApplyHalfGain(AudioFrame* frame);
+
+  static int Scale(float left, float right, AudioFrame* frame);
+
+  static int ScaleWithSat(float scale, AudioFrame* frame);
+};
+
+}  // namespace webrtc
+
+#endif  // AUDIO_UTILITY_AUDIO_FRAME_OPERATIONS_H_

TMessagesProj/jni/voip/webrtc/audio/utility/channel_mixer.cc

@@ -0,0 +1,99 @@
+/*
+ *  Copyright (c) 2012 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 "audio/utility/channel_mixer.h"
+
+#include "audio/utility/channel_mixing_matrix.h"
+#include "rtc_base/checks.h"
+#include "rtc_base/logging.h"
+#include "rtc_base/numerics/safe_conversions.h"
+
+namespace webrtc {
+
+ChannelMixer::ChannelMixer(ChannelLayout input_layout,
+                           ChannelLayout output_layout)
+    : input_layout_(input_layout),
+      output_layout_(output_layout),
+      input_channels_(ChannelLayoutToChannelCount(input_layout)),
+      output_channels_(ChannelLayoutToChannelCount(output_layout)) {
+  // Create the transformation matrix.
+  ChannelMixingMatrix matrix_builder(input_layout_, input_channels_,
+                                     output_layout_, output_channels_);
+  remapping_ = matrix_builder.CreateTransformationMatrix(&matrix_);
+}
+
+ChannelMixer::~ChannelMixer() = default;
+
+void ChannelMixer::Transform(AudioFrame* frame) {
+  RTC_DCHECK(frame);
+  RTC_DCHECK_EQ(matrix_[0].size(), static_cast<size_t>(input_channels_));
+  RTC_DCHECK_EQ(matrix_.size(), static_cast<size_t>(output_channels_));
+
+  // Leave the audio frame intact if the channel layouts for in and out are
+  // identical.
+  if (input_layout_ == output_layout_) {
+    return;
+  }
+
+  if (IsUpMixing()) {
+    RTC_CHECK_LE(frame->samples_per_channel() * output_channels_,
+                 frame->max_16bit_samples());
+  }
+
+  // Only change the number of output channels if the audio frame is muted.
+  if (frame->muted()) {
+    frame->num_channels_ = output_channels_;
+    frame->channel_layout_ = output_layout_;
+    return;
+  }
+
+  const int16_t* in_audio = frame->data();
+
+  // Only allocate fresh memory at first access or if the required size has
+  // increased.
+  // TODO(henrika): we might be able to do downmixing in-place and thereby avoid
+  // extra memory allocation and a memcpy.
+  const size_t num_elements = frame->samples_per_channel() * output_channels_;
+  if (audio_vector_ == nullptr || num_elements > audio_vector_size_) {
+    audio_vector_.reset(new int16_t[num_elements]);
+    audio_vector_size_ = num_elements;
+  }
+  int16_t* out_audio = audio_vector_.get();
+
+  // Modify the number of channels by creating a weighted sum of input samples
+  // where the weights (scale factors) for each output sample are given by the
+  // transformation matrix.
+  for (size_t i = 0; i < frame->samples_per_channel(); i++) {
+    for (size_t output_ch = 0; output_ch < output_channels_; ++output_ch) {
+      float acc_value = 0.0f;
+      for (size_t input_ch = 0; input_ch < input_channels_; ++input_ch) {
+        const float scale = matrix_[output_ch][input_ch];
+        // Scale should always be positive.
+        RTC_DCHECK_GE(scale, 0);
+        // Each output sample is a weighted sum of input samples.
+        acc_value += scale * in_audio[i * input_channels_ + input_ch];
+      }
+      const size_t index = output_channels_ * i + output_ch;
+      RTC_CHECK_LE(index, audio_vector_size_);
+      out_audio[index] = rtc::saturated_cast<int16_t>(acc_value);
+    }
+  }
+
+  // Update channel information.
+  frame->num_channels_ = output_channels_;
+  frame->channel_layout_ = output_layout_;
+
+  // Copy the output result to the audio frame in `frame`.
+  memcpy(
+      frame->mutable_data(), out_audio,
+      sizeof(int16_t) * frame->samples_per_channel() * frame->num_channels());
+}
+
+}  // namespace webrtc

TMessagesProj/jni/voip/webrtc/audio/utility/channel_mixer.h

@@ -0,0 +1,86 @@
+/*
+ *  Copyright (c) 2019 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.
+ */
+
+#ifndef AUDIO_UTILITY_CHANNEL_MIXER_H_
+#define AUDIO_UTILITY_CHANNEL_MIXER_H_
+
+#include <stddef.h>
+#include <stdint.h>
+
+#include <memory>
+#include <vector>
+
+#include "api/audio/audio_frame.h"
+#include "api/audio/channel_layout.h"
+
+namespace webrtc {
+
+// ChannelMixer is for converting audio between channel layouts.  The conversion
+// matrix is built upon construction and used during each Transform() call.  The
+// algorithm works by generating a conversion matrix mapping each output channel
+// to list of input channels.  The transform renders all of the output channels,
+// with each output channel rendered according to a weighted sum of the relevant
+// input channels as defined in the matrix.
+// This file is derived from Chromium's media/base/channel_mixer.h.
+class ChannelMixer {
+ public:
+  // To mix two channels into one and preserve loudness, we must apply
+  // (1 / sqrt(2)) gain to each.
+  static constexpr float kHalfPower = 0.707106781186547524401f;
+
+  ChannelMixer(ChannelLayout input_layout, ChannelLayout output_layout);
+  ~ChannelMixer();
+
+  // Transforms all input channels corresponding to the selected `input_layout`
+  // to the number of channels in the selected `output_layout`.
+  // Example usage (downmix from stereo to mono):
+  //
+  //   ChannelMixer mixer(CHANNEL_LAYOUT_STEREO, CHANNEL_LAYOUT_MONO);
+  //   AudioFrame frame;
+  //   frame.samples_per_channel_ = 160;
+  //   frame.num_channels_ = 2;
+  //   EXPECT_EQ(2u, frame.channels());
+  //   mixer.Transform(&frame);
+  //   EXPECT_EQ(1u, frame.channels());
+  //
+  void Transform(AudioFrame* frame);
+
+ private:
+  bool IsUpMixing() const { return output_channels_ > input_channels_; }
+
+  // Selected channel layouts.
+  const ChannelLayout input_layout_;
+  const ChannelLayout output_layout_;
+
+  // Channel counts for input and output.
+  const size_t input_channels_;
+  const size_t output_channels_;
+
+  // 2D matrix of output channels to input channels.
+  std::vector<std::vector<float> > matrix_;
+
+  // 1D array used as temporary storage during the transformation.
+  std::unique_ptr<int16_t[]> audio_vector_;
+
+  // Number of elements allocated for `audio_vector_`.
+  size_t audio_vector_size_ = 0;
+
+  // Optimization case for when we can simply remap the input channels to output
+  // channels, i.e., when all scaling factors in `matrix_` equals 1.0.
+  bool remapping_;
+
+  // Delete the copy constructor and assignment operator.
+  ChannelMixer(const ChannelMixer& other) = delete;
+  ChannelMixer& operator=(const ChannelMixer& other) = delete;
+};
+
+}  // namespace webrtc
+
+#endif  // AUDIO_UTILITY_CHANNEL_MIXER_H_

TMessagesProj/jni/voip/webrtc/audio/utility/channel_mixing_matrix.cc

@@ -0,0 +1,333 @@
+/*
+ *  Copyright (c) 2019 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 "audio/utility/channel_mixing_matrix.h"
+
+#include <stddef.h>
+
+#include <algorithm>
+
+#include "audio/utility/channel_mixer.h"
+#include "rtc_base/checks.h"
+#include "rtc_base/logging.h"
+#include "system_wrappers/include/field_trial.h"
+
+namespace webrtc {
+
+namespace {
+
+// Selects the default usage of VoIP channel mapping adjustments.
+bool UseChannelMappingAdjustmentsByDefault() {
+  return !field_trial::IsEnabled(
+      "WebRTC-VoIPChannelRemixingAdjustmentKillSwitch");
+}
+
+}  // namespace
+
+static void ValidateLayout(ChannelLayout layout) {
+  RTC_CHECK_NE(layout, CHANNEL_LAYOUT_NONE);
+  RTC_CHECK_LE(layout, CHANNEL_LAYOUT_MAX);
+  RTC_CHECK_NE(layout, CHANNEL_LAYOUT_UNSUPPORTED);
+  RTC_CHECK_NE(layout, CHANNEL_LAYOUT_DISCRETE);
+  RTC_CHECK_NE(layout, CHANNEL_LAYOUT_STEREO_AND_KEYBOARD_MIC);
+
+  // Verify there's at least one channel.  Should always be true here by virtue
+  // of not being one of the invalid layouts, but lets double check to be sure.
+  int channel_count = ChannelLayoutToChannelCount(layout);
+  RTC_DCHECK_GT(channel_count, 0);
+
+  // If we have more than one channel, verify a symmetric layout for sanity.
+  // The unit test will verify all possible layouts, so this can be a DCHECK.
+  // Symmetry allows simplifying the matrix building code by allowing us to
+  // assume that if one channel of a pair exists, the other will too.
+  if (channel_count > 1) {
+    // Assert that LEFT exists if and only if RIGHT exists, and so on.
+    RTC_DCHECK_EQ(ChannelOrder(layout, LEFT) >= 0,
+                  ChannelOrder(layout, RIGHT) >= 0);
+    RTC_DCHECK_EQ(ChannelOrder(layout, SIDE_LEFT) >= 0,
+                  ChannelOrder(layout, SIDE_RIGHT) >= 0);
+    RTC_DCHECK_EQ(ChannelOrder(layout, BACK_LEFT) >= 0,
+                  ChannelOrder(layout, BACK_RIGHT) >= 0);
+    RTC_DCHECK_EQ(ChannelOrder(layout, LEFT_OF_CENTER) >= 0,
+                  ChannelOrder(layout, RIGHT_OF_CENTER) >= 0);
+  } else {
+    RTC_DCHECK_EQ(layout, CHANNEL_LAYOUT_MONO);
+  }
+}
+
+ChannelMixingMatrix::ChannelMixingMatrix(ChannelLayout input_layout,
+                                         int input_channels,
+                                         ChannelLayout output_layout,
+                                         int output_channels)
+    : use_voip_channel_mapping_adjustments_(
+          UseChannelMappingAdjustmentsByDefault()),
+      input_layout_(input_layout),
+      input_channels_(input_channels),
+      output_layout_(output_layout),
+      output_channels_(output_channels) {
+  // Stereo down mix should never be the output layout.
+  RTC_CHECK_NE(output_layout, CHANNEL_LAYOUT_STEREO_DOWNMIX);
+
+  // Verify that the layouts are supported
+  if (input_layout != CHANNEL_LAYOUT_DISCRETE)
+    ValidateLayout(input_layout);
+  if (output_layout != CHANNEL_LAYOUT_DISCRETE)
+    ValidateLayout(output_layout);
+
+  // Special case for 5.0, 5.1 with back channels when upmixed to 7.0, 7.1,
+  // which should map the back LR to side LR.
+  if (input_layout_ == CHANNEL_LAYOUT_5_0_BACK &&
+      output_layout_ == CHANNEL_LAYOUT_7_0) {
+    input_layout_ = CHANNEL_LAYOUT_5_0;
+  } else if (input_layout_ == CHANNEL_LAYOUT_5_1_BACK &&
+             output_layout_ == CHANNEL_LAYOUT_7_1) {
+    input_layout_ = CHANNEL_LAYOUT_5_1;
+  }
+}
+
+ChannelMixingMatrix::~ChannelMixingMatrix() = default;
+
+bool ChannelMixingMatrix::CreateTransformationMatrix(
+    std::vector<std::vector<float>>* matrix) {
+  matrix_ = matrix;
+
+  // Size out the initial matrix.
+  matrix_->reserve(output_channels_);
+  for (int output_ch = 0; output_ch < output_channels_; ++output_ch)
+    matrix_->push_back(std::vector<float>(input_channels_, 0));
+
+  // First check for discrete case.
+  if (input_layout_ == CHANNEL_LAYOUT_DISCRETE ||
+      output_layout_ == CHANNEL_LAYOUT_DISCRETE) {
+    // If the number of input channels is more than output channels, then
+    // copy as many as we can then drop the remaining input channels.
+    // If the number of input channels is less than output channels, then
+    // copy them all, then zero out the remaining output channels.
+    int passthrough_channels = std::min(input_channels_, output_channels_);
+    for (int i = 0; i < passthrough_channels; ++i)
+      (*matrix_)[i][i] = 1;
+
+    return true;
+  }
+
+  // If specified, use adjusted channel mapping for the VoIP scenario.
+  if (use_voip_channel_mapping_adjustments_ &&
+      input_layout_ == CHANNEL_LAYOUT_MONO &&
+      ChannelLayoutToChannelCount(output_layout_) >= 2) {
+    // Only place the mono input in the front left and right channels.
+    (*matrix_)[0][0] = 1.f;
+    (*matrix_)[1][0] = 1.f;
+
+    for (size_t output_ch = 2; output_ch < matrix_->size(); ++output_ch) {
+      (*matrix_)[output_ch][0] = 0.f;
+    }
+    return true;
+  }
+
+  // Route matching channels and figure out which ones aren't accounted for.
+  for (Channels ch = LEFT; ch < CHANNELS_MAX + 1;
+       ch = static_cast<Channels>(ch + 1)) {
+    int input_ch_index = ChannelOrder(input_layout_, ch);
+    if (input_ch_index < 0)
+      continue;
+
+    int output_ch_index = ChannelOrder(output_layout_, ch);
+    if (output_ch_index < 0) {
+      unaccounted_inputs_.push_back(ch);
+      continue;
+    }
+
+    RTC_DCHECK_LT(static_cast<size_t>(output_ch_index), matrix_->size());
+    RTC_DCHECK_LT(static_cast<size_t>(input_ch_index),
+                  (*matrix_)[output_ch_index].size());
+    (*matrix_)[output_ch_index][input_ch_index] = 1;
+  }
+
+  // If all input channels are accounted for, there's nothing left to do.
+  if (unaccounted_inputs_.empty()) {
+    // Since all output channels map directly to inputs we can optimize.
+    return true;
+  }
+
+  // Mix front LR into center.
+  if (IsUnaccounted(LEFT)) {
+    // When down mixing to mono from stereo, we need to be careful of full scale
+    // stereo mixes.  Scaling by 1 / sqrt(2) here will likely lead to clipping
+    // so we use 1 / 2 instead.
+    float scale =
+        (output_layout_ == CHANNEL_LAYOUT_MONO && input_channels_ == 2)
+            ? 0.5
+            : ChannelMixer::kHalfPower;
+    Mix(LEFT, CENTER, scale);
+    Mix(RIGHT, CENTER, scale);
+  }
+
+  // Mix center into front LR.
+  if (IsUnaccounted(CENTER)) {
+    // When up mixing from mono, just do a copy to front LR.
+    float scale =
+        (input_layout_ == CHANNEL_LAYOUT_MONO) ? 1 : ChannelMixer::kHalfPower;
+    MixWithoutAccounting(CENTER, LEFT, scale);
+    Mix(CENTER, RIGHT, scale);
+  }
+
+  // Mix back LR into: side LR || back center || front LR || front center.
+  if (IsUnaccounted(BACK_LEFT)) {
+    if (HasOutputChannel(SIDE_LEFT)) {
+      // If the input has side LR, mix back LR into side LR, but instead if the
+      // input doesn't have side LR (but output does) copy back LR to side LR.
+      float scale = HasInputChannel(SIDE_LEFT) ? ChannelMixer::kHalfPower : 1;
+      Mix(BACK_LEFT, SIDE_LEFT, scale);
+      Mix(BACK_RIGHT, SIDE_RIGHT, scale);
+    } else if (HasOutputChannel(BACK_CENTER)) {
+      // Mix back LR into back center.
+      Mix(BACK_LEFT, BACK_CENTER, ChannelMixer::kHalfPower);
+      Mix(BACK_RIGHT, BACK_CENTER, ChannelMixer::kHalfPower);
+    } else if (output_layout_ > CHANNEL_LAYOUT_MONO) {
+      // Mix back LR into front LR.
+      Mix(BACK_LEFT, LEFT, ChannelMixer::kHalfPower);
+      Mix(BACK_RIGHT, RIGHT, ChannelMixer::kHalfPower);
+    } else {
+      // Mix back LR into front center.
+      Mix(BACK_LEFT, CENTER, ChannelMixer::kHalfPower);
+      Mix(BACK_RIGHT, CENTER, ChannelMixer::kHalfPower);
+    }
+  }
+
+  // Mix side LR into: back LR || back center || front LR || front center.
+  if (IsUnaccounted(SIDE_LEFT)) {
+    if (HasOutputChannel(BACK_LEFT)) {
+      // If the input has back LR, mix side LR into back LR, but instead if the
+      // input doesn't have back LR (but output does) copy side LR to back LR.
+      float scale = HasInputChannel(BACK_LEFT) ? ChannelMixer::kHalfPower : 1;
+      Mix(SIDE_LEFT, BACK_LEFT, scale);
+      Mix(SIDE_RIGHT, BACK_RIGHT, scale);
+    } else if (HasOutputChannel(BACK_CENTER)) {
+      // Mix side LR into back center.
+      Mix(SIDE_LEFT, BACK_CENTER, ChannelMixer::kHalfPower);
+      Mix(SIDE_RIGHT, BACK_CENTER, ChannelMixer::kHalfPower);
+    } else if (output_layout_ > CHANNEL_LAYOUT_MONO) {
+      // Mix side LR into front LR.
+      Mix(SIDE_LEFT, LEFT, ChannelMixer::kHalfPower);
+      Mix(SIDE_RIGHT, RIGHT, ChannelMixer::kHalfPower);
+    } else {
+      // Mix side LR into front center.
+      Mix(SIDE_LEFT, CENTER, ChannelMixer::kHalfPower);
+      Mix(SIDE_RIGHT, CENTER, ChannelMixer::kHalfPower);
+    }
+  }
+
+  // Mix back center into: back LR || side LR || front LR || front center.
+  if (IsUnaccounted(BACK_CENTER)) {
+    if (HasOutputChannel(BACK_LEFT)) {
+      // Mix back center into back LR.
+      MixWithoutAccounting(BACK_CENTER, BACK_LEFT, ChannelMixer::kHalfPower);
+      Mix(BACK_CENTER, BACK_RIGHT, ChannelMixer::kHalfPower);
+    } else if (HasOutputChannel(SIDE_LEFT)) {
+      // Mix back center into side LR.
+      MixWithoutAccounting(BACK_CENTER, SIDE_LEFT, ChannelMixer::kHalfPower);
+      Mix(BACK_CENTER, SIDE_RIGHT, ChannelMixer::kHalfPower);
+    } else if (output_layout_ > CHANNEL_LAYOUT_MONO) {
+      // Mix back center into front LR.
+      // TODO(dalecurtis): Not sure about these values?
+      MixWithoutAccounting(BACK_CENTER, LEFT, ChannelMixer::kHalfPower);
+      Mix(BACK_CENTER, RIGHT, ChannelMixer::kHalfPower);
+    } else {
+      // Mix back center into front center.
+      // TODO(dalecurtis): Not sure about these values?
+      Mix(BACK_CENTER, CENTER, ChannelMixer::kHalfPower);
+    }
+  }
+
+  // Mix LR of center into: front LR || front center.
+  if (IsUnaccounted(LEFT_OF_CENTER)) {
+    if (HasOutputChannel(LEFT)) {
+      // Mix LR of center into front LR.
+      Mix(LEFT_OF_CENTER, LEFT, ChannelMixer::kHalfPower);
+      Mix(RIGHT_OF_CENTER, RIGHT, ChannelMixer::kHalfPower);
+    } else {
+      // Mix LR of center into front center.
+      Mix(LEFT_OF_CENTER, CENTER, ChannelMixer::kHalfPower);
+      Mix(RIGHT_OF_CENTER, CENTER, ChannelMixer::kHalfPower);
+    }
+  }
+
+  // Mix LFE into: front center || front LR.
+  if (IsUnaccounted(LFE)) {
+    if (!HasOutputChannel(CENTER)) {
+      // Mix LFE into front LR.
+      MixWithoutAccounting(LFE, LEFT, ChannelMixer::kHalfPower);
+      Mix(LFE, RIGHT, ChannelMixer::kHalfPower);
+    } else {
+      // Mix LFE into front center.
+      Mix(LFE, CENTER, ChannelMixer::kHalfPower);
+    }
+  }
+
+  // All channels should now be accounted for.
+  RTC_DCHECK(unaccounted_inputs_.empty());
+
+  // See if the output `matrix_` is simply a remapping matrix.  If each input
+  // channel maps to a single output channel we can simply remap.  Doing this
+  // programmatically is less fragile than logic checks on channel mappings.
+  for (int output_ch = 0; output_ch < output_channels_; ++output_ch) {
+    int input_mappings = 0;
+    for (int input_ch = 0; input_ch < input_channels_; ++input_ch) {
+      // We can only remap if each row contains a single scale of 1.  I.e., each
+      // output channel is mapped from a single unscaled input channel.
+      if ((*matrix_)[output_ch][input_ch] != 1 || ++input_mappings > 1)
+        return false;
+    }
+  }
+
+  // If we've gotten here, `matrix_` is simply a remapping.
+  return true;
+}
+
+void ChannelMixingMatrix::AccountFor(Channels ch) {
+  unaccounted_inputs_.erase(
+      std::find(unaccounted_inputs_.begin(), unaccounted_inputs_.end(), ch));
+}
+
+bool ChannelMixingMatrix::IsUnaccounted(Channels ch) const {
+  return std::find(unaccounted_inputs_.begin(), unaccounted_inputs_.end(),
+                   ch) != unaccounted_inputs_.end();
+}
+
+bool ChannelMixingMatrix::HasInputChannel(Channels ch) const {
+  return ChannelOrder(input_layout_, ch) >= 0;
+}
+
+bool ChannelMixingMatrix::HasOutputChannel(Channels ch) const {
+  return ChannelOrder(output_layout_, ch) >= 0;
+}
+
+void ChannelMixingMatrix::Mix(Channels input_ch,
+                              Channels output_ch,
+                              float scale) {
+  MixWithoutAccounting(input_ch, output_ch, scale);
+  AccountFor(input_ch);
+}
+
+void ChannelMixingMatrix::MixWithoutAccounting(Channels input_ch,
+                                               Channels output_ch,
+                                               float scale) {
+  int input_ch_index = ChannelOrder(input_layout_, input_ch);
+  int output_ch_index = ChannelOrder(output_layout_, output_ch);
+
+  RTC_DCHECK(IsUnaccounted(input_ch));
+  RTC_DCHECK_GE(input_ch_index, 0);
+  RTC_DCHECK_GE(output_ch_index, 0);
+
+  RTC_DCHECK_EQ((*matrix_)[output_ch_index][input_ch_index], 0);
+  (*matrix_)[output_ch_index][input_ch_index] = scale;
+}
+
+}  // namespace webrtc

TMessagesProj/jni/voip/webrtc/audio/utility/channel_mixing_matrix.h

@@ -0,0 +1,76 @@
+/*
+ *  Copyright (c) 2019 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.
+ */
+
+#ifndef AUDIO_UTILITY_CHANNEL_MIXING_MATRIX_H_
+#define AUDIO_UTILITY_CHANNEL_MIXING_MATRIX_H_
+
+#include <vector>
+
+#include "api/audio/channel_layout.h"
+
+namespace webrtc {
+
+class ChannelMixingMatrix {
+ public:
+  ChannelMixingMatrix(ChannelLayout input_layout,
+                      int input_channels,
+                      ChannelLayout output_layout,
+                      int output_channels);
+
+  ~ChannelMixingMatrix();
+
+  // Create the transformation matrix of input channels to output channels.
+  // Updates the empty matrix with the transformation, and returns true
+  // if the transformation is just a remapping of channels (no mixing).
+  // The size of `matrix` is `output_channels` x `input_channels`, i.e., the
+  // number of rows equals the number of output channels and the number of
+  // columns corresponds to the number of input channels.
+  // This file is derived from Chromium's media/base/channel_mixing_matrix.h.
+  bool CreateTransformationMatrix(std::vector<std::vector<float>>* matrix);
+
+ private:
+  const bool use_voip_channel_mapping_adjustments_;
+
+  // Result transformation of input channels to output channels
+  std::vector<std::vector<float>>* matrix_;
+
+  // Input and output channel layout provided during construction.
+  ChannelLayout input_layout_;
+  int input_channels_;
+  ChannelLayout output_layout_;
+  int output_channels_;
+
+  // Helper variable for tracking which inputs are currently unaccounted,
+  // should be empty after construction completes.
+  std::vector<Channels> unaccounted_inputs_;
+
+  // Helper methods for managing unaccounted input channels.
+  void AccountFor(Channels ch);
+  bool IsUnaccounted(Channels ch) const;
+
+  // Helper methods for checking if `ch` exists in either `input_layout_` or
+  // `output_layout_` respectively.
+  bool HasInputChannel(Channels ch) const;
+  bool HasOutputChannel(Channels ch) const;
+
+  // Helper methods for updating `matrix_` with the proper value for
+  // mixing `input_ch` into `output_ch`.  MixWithoutAccounting() does not
+  // remove the channel from `unaccounted_inputs_`.
+  void Mix(Channels input_ch, Channels output_ch, float scale);
+  void MixWithoutAccounting(Channels input_ch, Channels output_ch, float scale);
+
+  // Delete the copy constructor and assignment operator.
+  ChannelMixingMatrix(const ChannelMixingMatrix& other) = delete;
+  ChannelMixingMatrix& operator=(const ChannelMixingMatrix& other) = delete;
+};
+
+}  // namespace webrtc
+
+#endif  // AUDIO_UTILITY_CHANNEL_MIXING_MATRIX_H_