remoting/codec/audio_encoder_opus.cc

   1 // Copyright (c) 2012 The Chromium Authors. All rights reserved.
   2 // Use of this source code is governed by a BSD-style license that can be
   3 // found in the LICENSE file.
   4
   5 #include "remoting/codec/audio_encoder_opus.h"
   6
   7 #include "base/bind.h"
   8 #include "base/logging.h"
   9 #include "base/time/time.h"
  10 #include "media/base/audio_bus.h"
  11 #include "media/base/multi_channel_resampler.h"
  12 #include "third_party/opus/src/include/opus.h"
  13
  14 namespace remoting {
  15
  16 namespace {
  17
  18 // Output 160 kb/s bitrate.
  19 const int kOutputBitrateBps = 160 * 1024;
  20
  21 // Opus doesn't support 44100 sampling rate so we always resample to 48kHz.
  22 const AudioPacket::SamplingRate kOpusSamplingRate =
  23     AudioPacket::SAMPLING_RATE_48000;
  24
  25 // Opus supports frame sizes of 2.5, 5, 10, 20, 40 and 60 ms. We use 20 ms
  26 // frames to balance latency and efficiency.
  27 const int kFrameSizeMs = 20;
  28
  29 // Number of samples per frame when using default sampling rate.
  30 const int kFrameSamples =
  31     kOpusSamplingRate * kFrameSizeMs / base::Time::kMillisecondsPerSecond;
  32
  33 const AudioPacket::BytesPerSample kBytesPerSample =
  34     AudioPacket::BYTES_PER_SAMPLE_2;
  35
  36 bool IsSupportedSampleRate(int rate) {
  37   return rate == 44100 || rate == 48000;
  38 }
  39
  40 }  // namespace
  41
  42 AudioEncoderOpus::AudioEncoderOpus()
  43     : sampling_rate_(0),
  44       channels_(AudioPacket::CHANNELS_STEREO),
  45       encoder_(NULL),
  46       frame_size_(0),
  47       resampling_data_(NULL),
  48       resampling_data_size_(0),
  49       resampling_data_pos_(0) {
  50 }
  51
  52 AudioEncoderOpus::~AudioEncoderOpus() {
  53   DestroyEncoder();
  54 }
  55
  56 void AudioEncoderOpus::InitEncoder() {
  57   DCHECK(!encoder_);
  58   int error;
  59   encoder_ = opus_encoder_create(kOpusSamplingRate, channels_,
  60                                  OPUS_APPLICATION_AUDIO, &error);
  61   if (!encoder_) {
  62     LOG(ERROR) << "Failed to create OPUS encoder. Error code: " << error;
  63     return;
  64   }
  65
  66   opus_encoder_ctl(encoder_, OPUS_SET_BITRATE(kOutputBitrateBps));
  67
  68   frame_size_ = sampling_rate_ * kFrameSizeMs /
  69       base::Time::kMillisecondsPerSecond;
  70
  71   if (sampling_rate_ != kOpusSamplingRate) {
  72     resample_buffer_.reset(
  73         new char[kFrameSamples * kBytesPerSample * channels_]);
  74     // TODO(sergeyu): Figure out the right buffer size to use per packet instead
  75     // of using media::SincResampler::kDefaultRequestSize.
  76     resampler_.reset(new media::MultiChannelResampler(
  77         channels_,
  78         static_cast<double>(sampling_rate_) / kOpusSamplingRate,
  79         media::SincResampler::kDefaultRequestSize,
  80         base::Bind(&AudioEncoderOpus::FetchBytesToResample,
  81                    base::Unretained(this))));
  82     resampler_bus_ = media::AudioBus::Create(channels_, kFrameSamples);
  83   }
  84
  85   // Drop leftover data because it's for different sampling rate.
  86   leftover_samples_ = 0;
  87   leftover_buffer_size_ =
  88       frame_size_ + media::SincResampler::kDefaultRequestSize;
  89   leftover_buffer_.reset(
  90       new int16[leftover_buffer_size_ * channels_]);
  91 }
  92
  93 void AudioEncoderOpus::DestroyEncoder() {
  94   if (encoder_) {
  95     opus_encoder_destroy(encoder_);
  96     encoder_ = NULL;
  97   }
  98
  99   resampler_.reset();
 100 }
 101
 102 bool AudioEncoderOpus::ResetForPacket(AudioPacket* packet) {
 103   if (packet->channels() != channels_ ||
 104       packet->sampling_rate() != sampling_rate_) {
 105     DestroyEncoder();
 106
 107     channels_ = packet->channels();
 108     sampling_rate_ = packet->sampling_rate();
 109
 110     if (channels_ <= 0 || channels_ > 2 ||
 111         !IsSupportedSampleRate(sampling_rate_)) {
 112       LOG(WARNING) << "Unsupported OPUS parameters: "
 113                    << channels_ << " channels with "
 114                    << sampling_rate_ << " samples per second.";
 115       return false;
 116     }
 117
 118     InitEncoder();
 119   }
 120
 121   return encoder_ != NULL;
 122 }
 123
 124 void AudioEncoderOpus::FetchBytesToResample(int resampler_frame_delay,
 125                                             media::AudioBus* audio_bus) {
 126   DCHECK(resampling_data_);
 127   int samples_left = (resampling_data_size_ - resampling_data_pos_) /
 128       kBytesPerSample / channels_;
 129   DCHECK_LE(audio_bus->frames(), samples_left);
 130   audio_bus->FromInterleaved(
 131       resampling_data_ + resampling_data_pos_,
 132       audio_bus->frames(), kBytesPerSample);
 133   resampling_data_pos_ += audio_bus->frames() * kBytesPerSample * channels_;
 134   DCHECK_LE(resampling_data_pos_, static_cast<int>(resampling_data_size_));
 135 }
 136
 137 scoped_ptr<AudioPacket> AudioEncoderOpus::Encode(
 138     scoped_ptr<AudioPacket> packet) {
 139   DCHECK_EQ(AudioPacket::ENCODING_RAW, packet->encoding());
 140   DCHECK_EQ(1, packet->data_size());
 141   DCHECK_EQ(kBytesPerSample, packet->bytes_per_sample());
 142
 143   if (!ResetForPacket(packet.get())) {
 144     LOG(ERROR) << "Encoder initialization failed";
 145     return scoped_ptr<AudioPacket>();
 146   }
 147
 148   int samples_in_packet = packet->data(0).size() / kBytesPerSample / channels_;
 149   const int16* next_sample =
 150       reinterpret_cast<const int16*>(packet->data(0).data());
 151
 152   // Create a new packet of encoded data.
 153   scoped_ptr<AudioPacket> encoded_packet(new AudioPacket());
 154   encoded_packet->set_encoding(AudioPacket::ENCODING_OPUS);
 155   encoded_packet->set_sampling_rate(kOpusSamplingRate);
 156   encoded_packet->set_channels(channels_);
 157
 158   int prefetch_samples =
 159       resampler_.get() ? media::SincResampler::kDefaultRequestSize : 0;
 160   int samples_wanted = frame_size_ + prefetch_samples;
 161
 162   while (leftover_samples_ + samples_in_packet >= samples_wanted) {
 163     const int16* pcm_buffer = NULL;
 164
 165     // Combine the packet with the leftover samples, if any.
 166     if (leftover_samples_ > 0) {
 167       pcm_buffer = leftover_buffer_.get();
 168       int samples_to_copy = samples_wanted - leftover_samples_;
 169       memcpy(leftover_buffer_.get() + leftover_samples_ * channels_,
 170              next_sample, samples_to_copy * kBytesPerSample * channels_);
 171     } else {
 172       pcm_buffer = next_sample;
 173     }
 174
 175     // Resample data if necessary.
 176     int samples_consumed = 0;
 177     if (resampler_.get()) {
 178       resampling_data_ = reinterpret_cast<const char*>(pcm_buffer);
 179       resampling_data_pos_ = 0;
 180       resampling_data_size_ = samples_wanted * channels_ * kBytesPerSample;
 181       resampler_->Resample(kFrameSamples, resampler_bus_.get());
 182       resampling_data_ = NULL;
 183       samples_consumed = resampling_data_pos_ / channels_ / kBytesPerSample;
 184
 185       resampler_bus_->ToInterleaved(kFrameSamples, kBytesPerSample,
 186                                     resample_buffer_.get());
 187       pcm_buffer = reinterpret_cast<int16*>(resample_buffer_.get());
 188     } else {
 189       samples_consumed = frame_size_;
 190     }
 191
 192     // Initialize output buffer.
 193     std::string* data = encoded_packet->add_data();
 194     data->resize(kFrameSamples * kBytesPerSample * channels_);
 195
 196     // Encode.
 197     unsigned char* buffer =
 198         reinterpret_cast<unsigned char*>(string_as_array(data));
 199     int result = opus_encode(encoder_, pcm_buffer, kFrameSamples,
 200                              buffer, data->length());
 201     if (result < 0) {
 202       LOG(ERROR) << "opus_encode() failed with error code: " << result;
 203       return scoped_ptr<AudioPacket>();
 204     }
 205
 206     DCHECK_LE(result, static_cast<int>(data->length()));
 207     data->resize(result);
 208
 209     // Cleanup leftover buffer.
 210     if (samples_consumed >= leftover_samples_) {
 211       samples_consumed -= leftover_samples_;
 212       leftover_samples_ = 0;
 213       next_sample += samples_consumed * channels_;
 214       samples_in_packet -= samples_consumed;
 215     } else {
 216       leftover_samples_ -= samples_consumed;
 217       memmove(leftover_buffer_.get(),
 218               leftover_buffer_.get() + samples_consumed * channels_,
 219               leftover_samples_ * channels_ * kBytesPerSample);
 220     }
 221   }
 222
 223   // Store the leftover samples.
 224   if (samples_in_packet > 0) {
 225     DCHECK_LE(leftover_samples_ + samples_in_packet, leftover_buffer_size_);
 226     memmove(leftover_buffer_.get() + leftover_samples_ * channels_,
 227             next_sample, samples_in_packet * kBytesPerSample * channels_);
 228     leftover_samples_ += samples_in_packet;
 229   }
 230
 231   // Return NULL if there's nothing in the packet.
 232   if (encoded_packet->data_size() == 0)
 233     return scoped_ptr<AudioPacket>();
 234
 235   return encoded_packet.Pass();
 236 }
 237
 238 }  // namespace remoting