C++ AudioSegment类代码示例

void
MediaEngineWebRTCAudioSource::Process(const int channel,
  const webrtc::ProcessingTypes type, sample* audio10ms,
  const int length, const int samplingFreq, const bool isStereo)
{
  ReentrantMonitorAutoEnter enter(mMonitor);
  if (mState != kStarted)
    return;

  uint32_t len = mSources.Length();
  for (uint32_t i = 0; i < len; i++) {
    nsRefPtr<SharedBuffer> buffer = SharedBuffer::Create(length * sizeof(sample));

    sample* dest = static_cast<sample*>(buffer->Data());
    memcpy(dest, audio10ms, length * sizeof(sample));

    AudioSegment segment;
    nsAutoTArray<const sample*,1> channels;
    channels.AppendElement(dest);
    segment.AppendFrames(buffer.forget(), channels, length);

    SourceMediaStream *source = mSources[i];
    if (source) {
      // This is safe from any thread, and is safe if the track is Finished
      // or Destroyed
      source->AppendToTrack(mTrackID, &segment);
    }
  }

  return;
}

NS_IMETHODIMP
MediaEngineDefaultAudioSource::Notify(nsITimer* aTimer)
{
  TimeStamp now = TimeStamp::Now();
  TimeDuration timeSinceLastNotify = now - mLastNotify;
  mLastNotify = now;
  TrackTicks samplesSinceLastNotify =
    RateConvertTicksRoundUp(AUDIO_RATE, 1000000, timeSinceLastNotify.ToMicroseconds());

  // If it's been longer since the last Notify() than mBufferSize holds, we
  // have underrun and the MSG had to append silence while waiting for us
  // to push more data. In this case we reset to mBufferSize again.
  TrackTicks samplesToAppend = std::min(samplesSinceLastNotify, mBufferSize);

  AudioSegment segment;
  AppendToSegment(segment, samplesToAppend);
  mSource->AppendToTrack(mTrackID, &segment);

  // Generate null data for fake tracks.
  if (mHasFakeTracks) {
    for (int i = 0; i < kFakeAudioTrackCount; ++i) {
      AudioSegment nullSegment;
      nullSegment.AppendNullData(samplesToAppend);
      mSource->AppendToTrack(kTrackCount + kFakeVideoTrackCount+i, &nullSegment);
    }
  }
  return NS_OK;
}

nsresult
MediaEngineDefaultAudioSource::Start(SourceMediaStream* aStream, TrackID aID)
{
    if (mState != kAllocated) {
        return NULL;
    }

    mTimer = do_CreateInstance(NS_TIMER_CONTRACTID);
    if (!mTimer) {
        return NULL;
    }

    mSource = aStream;

    // AddTrack will take ownership of segment
    AudioSegment* segment = new AudioSegment();
    segment->Init(CHANNELS);
    mSource->AddTrack(aID, RATE, 0, segment);

    // We aren't going to add any more tracks
    mSource->AdvanceKnownTracksTime(STREAM_TIME_MAX);

    // Remember TrackID so we can finish later
    mTrackID = aID;

    // 1 Audio frame per Video frame
    mTimer->InitWithCallback(this, 1000 / FPS, nsITimer::TYPE_REPEATING_SLACK);
    mState = kStarted;

    return NS_OK;
}

NS_IMETHODIMP
MediaEngineDefaultAudioSource::Notify(nsITimer* aTimer)
{
  AudioSegment segment;
  segment.InsertNullDataAtStart(AUDIO_RATE/100); // 10ms of fake data

  mSource->AppendToTrack(mTrackID, &segment);

  return NS_OK;
}

nsresult
MediaEngineDefaultAudioSource::Start(SourceMediaStream* aStream, TrackID aID,
                                     const PrincipalHandle& aPrincipalHandle)
{
  if (mState != kAllocated) {
    return NS_ERROR_FAILURE;
  }

  mTimer = do_CreateInstance(NS_TIMER_CONTRACTID);
  if (!mTimer) {
    return NS_ERROR_FAILURE;
  }

  mSource = aStream;

  // We try to keep the appended data at this size.
  // Make it two timer intervals to try to avoid underruns.
  mBufferSize = 2 * (AUDIO_RATE * DEFAULT_AUDIO_TIMER_MS) / 1000;

  // AddTrack will take ownership of segment
  AudioSegment* segment = new AudioSegment();
  AppendToSegment(*segment, mBufferSize);
  mSource->AddAudioTrack(aID, AUDIO_RATE, 0, segment, SourceMediaStream::ADDTRACK_QUEUED);

  if (mHasFakeTracks) {
    for (int i = 0; i < kFakeAudioTrackCount; ++i) {
      segment = new AudioSegment();
      segment->AppendNullData(mBufferSize);
      mSource->AddAudioTrack(kTrackCount + kFakeVideoTrackCount+i,
                             AUDIO_RATE, 0, segment, SourceMediaStream::ADDTRACK_QUEUED);
    }
  }

  // Remember TrackID so we can finish later
  mTrackID = aID;

  // Remember PrincipalHandle since we don't append in NotifyPull.
  mPrincipalHandle = aPrincipalHandle;

  mLastNotify = TimeStamp::Now();

  // 1 Audio frame per 10ms
#if defined(MOZ_WIDGET_GONK) && defined(DEBUG)
// B2G emulator debug is very, very slow and has problems dealing with realtime audio inputs
  mTimer->InitWithCallback(this, DEFAULT_AUDIO_TIMER_MS*10,
                           nsITimer::TYPE_REPEATING_PRECISE_CAN_SKIP);
#else
  mTimer->InitWithCallback(this, DEFAULT_AUDIO_TIMER_MS,
                           nsITimer::TYPE_REPEATING_PRECISE_CAN_SKIP);
#endif
  mState = kStarted;

  return NS_OK;
}

NS_IMETHODIMP
MediaEngineDefaultAudioSource::Notify(nsITimer* aTimer)
{
    AudioSegment segment;
    segment.Init(CHANNELS);
    segment.InsertNullDataAtStart(1);

    mSource->AppendToTrack(mTrackID, &segment);

    return NS_OK;
}

NS_IMETHODIMP
MediaEngineDefaultAudioSource::Notify(nsITimer* aTimer)
{
  AudioSegment segment;

  // Notify timer is set every DEFAULT_AUDIO_TIMER_MS milliseconds.
  segment.InsertNullDataAtStart((AUDIO_RATE * MediaEngine::DEFAULT_AUDIO_TIMER_MS) / 1000);

  mSource->AppendToTrack(mTrackID, &segment);

  return NS_OK;
}

void
AudioCaptureStream::ProcessInput(GraphTime aFrom, GraphTime aTo,
                                 uint32_t aFlags)
{
  uint32_t inputCount = mInputs.Length();
  StreamBuffer::Track* track = EnsureTrack(mTrackId);
  // Notify the DOM everything is in order.
  if (!mTrackCreated) {
    for (uint32_t i = 0; i < mListeners.Length(); i++) {
      MediaStreamListener* l = mListeners[i];
      AudioSegment tmp;
      l->NotifyQueuedTrackChanges(
        Graph(), mTrackId, 0, MediaStreamListener::TRACK_EVENT_CREATED, tmp);
      l->NotifyFinishedTrackCreation(Graph());
    }
    mTrackCreated = true;
  }

  // If the captured stream is connected back to a object on the page (be it an
  // HTMLMediaElement with a stream as source, or an AudioContext), a cycle
  // situation occur. This can work if it's an AudioContext with at least one
  // DelayNode, but the MSG will mute the whole cycle otherwise.
  if (mFinished || InMutedCycle() || inputCount == 0) {
    track->Get<AudioSegment>()->AppendNullData(aTo - aFrom);
  } else {
    // We mix down all the tracks of all inputs, to a stereo track. Everything
    // is {up,down}-mixed to stereo.
    mMixer.StartMixing();
    AudioSegment output;
    for (uint32_t i = 0; i < inputCount; i++) {
      MediaStream* s = mInputs[i]->GetSource();
      StreamBuffer::TrackIter tracks(s->GetStreamBuffer(), MediaSegment::AUDIO);
      while (!tracks.IsEnded()) {
        AudioSegment* inputSegment = tracks->Get<AudioSegment>();
        StreamTime inputStart = s->GraphTimeToStreamTimeWithBlocking(aFrom);
        StreamTime inputEnd = s->GraphTimeToStreamTimeWithBlocking(aTo);
        AudioSegment toMix;
        toMix.AppendSlice(*inputSegment, inputStart, inputEnd);
        // Care for streams blocked in the [aTo, aFrom] range.
        if (inputEnd - inputStart < aTo - aFrom) {
          toMix.AppendNullData((aTo - aFrom) - (inputEnd - inputStart));
        }
        toMix.Mix(mMixer, MONO, Graph()->GraphRate());
        tracks.Next();
      }
    }
    // This calls MixerCallback below
    mMixer.FinishMixing();
  }

  // Regardless of the status of the input tracks, we go foward.
  mBuffer.AdvanceKnownTracksTime(GraphTimeToStreamTimeWithBlocking((aTo)));
}

nsresult
OmxAudioTrackEncoder::GetEncodedTrack(EncodedFrameContainer& aData)
{
  PROFILER_LABEL("OmxAACAudioTrackEncoder", "GetEncodedTrack",
    js::ProfileEntry::Category::OTHER);
  AudioSegment segment;
  // Move all the samples from mRawSegment to segment. We only hold
  // the monitor in this block.
  {
    ReentrantMonitorAutoEnter mon(mReentrantMonitor);

    // Wait if mEncoder is not initialized nor canceled.
    while (!mInitialized && !mCanceled) {
      mReentrantMonitor.Wait();
    }

    if (mCanceled || mEncodingComplete) {
      return NS_ERROR_FAILURE;
    }

    segment.AppendFrom(&mRawSegment);
  }

  nsresult rv;
  if (segment.GetDuration() == 0) {
    // Notify EOS at least once, even if segment is empty.
    if (mEndOfStream && !mEosSetInEncoder) {
      mEosSetInEncoder = true;
      rv = mEncoder->Encode(segment, OMXCodecWrapper::BUFFER_EOS);
      NS_ENSURE_SUCCESS(rv, rv);
    }
    // Nothing to encode but encoder could still have encoded data for earlier
    // input.
    return AppendEncodedFrames(aData);
  }

  // OMX encoder has limited input buffers only so we have to feed input and get
  // output more than once if there are too many samples pending in segment.
  while (segment.GetDuration() > 0) {
    rv = mEncoder->Encode(segment,
                          mEndOfStream ? OMXCodecWrapper::BUFFER_EOS : 0);
    NS_ENSURE_SUCCESS(rv, rv);

    rv = AppendEncodedFrames(aData);
    NS_ENSURE_SUCCESS(rv, rv);
  }

  return NS_OK;
}

void
nsSpeechTask::SendAudioImpl(RefPtr<mozilla::SharedBuffer>& aSamples, uint32_t aDataLen)
{
  if (aDataLen == 0) {
    mStream->EndAllTrackAndFinish();
    return;
  }

  AudioSegment segment;
  nsAutoTArray<const int16_t*, 1> channelData;
  channelData.AppendElement(static_cast<int16_t*>(aSamples->Data()));
  segment.AppendFrames(aSamples.forget(), channelData, aDataLen);
  mStream->AppendToTrack(1, &segment);
  mStream->AdvanceKnownTracksTime(STREAM_TIME_MAX);
}

AudioSegment*
SpeechRecognition::CreateAudioSegment(nsTArray<nsRefPtr<SharedBuffer>>& aChunks)
{
  AudioSegment* segment = new AudioSegment();
  for (uint32_t i = 0; i < aChunks.Length(); ++i) {
    nsRefPtr<SharedBuffer> buffer = aChunks[i];
    const int16_t* chunkData = static_cast<const int16_t*>(buffer->Data());

    nsAutoTArray<const int16_t*, 1> channels;
    channels.AppendElement(chunkData);
    segment->AppendFrames(buffer.forget(), channels, mAudioSamplesPerChunk);
  }

  return segment;
}

NS_IMETHODIMP
MediaEngineDefaultAudioSource::Notify(nsITimer* aTimer)
{
  AudioSegment segment;
  nsRefPtr<SharedBuffer> buffer = SharedBuffer::Create(AUDIO_FRAME_LENGTH * sizeof(int16_t));
  int16_t* dest = static_cast<int16_t*>(buffer->Data());

  mSineGenerator->generate(dest, AUDIO_FRAME_LENGTH);
  nsAutoTArray<const int16_t*,1> channels;
  channels.AppendElement(dest);
  segment.AppendFrames(buffer.forget(), channels, AUDIO_FRAME_LENGTH);
  mSource->AppendToTrack(mTrackID, &segment);

  return NS_OK;
}

static void
SendStreamAudio(DecodedStreamData* aStream, int64_t aStartTime,
                MediaData* aData, AudioSegment* aOutput,
                uint32_t aRate, double aVolume)
{
  MOZ_ASSERT(aData);
  AudioData* audio = aData->As<AudioData>();
  // This logic has to mimic AudioSink closely to make sure we write
  // the exact same silences
  CheckedInt64 audioWrittenOffset = aStream->mAudioFramesWritten +
                                    UsecsToFrames(aStartTime, aRate);
  CheckedInt64 frameOffset = UsecsToFrames(audio->mTime, aRate);

  if (!audioWrittenOffset.isValid() ||
      !frameOffset.isValid() ||
      // ignore packet that we've already processed
      frameOffset.value() + audio->mFrames <= audioWrittenOffset.value()) {
    return;
  }

  if (audioWrittenOffset.value() < frameOffset.value()) {
    int64_t silentFrames = frameOffset.value() - audioWrittenOffset.value();
    // Write silence to catch up
    AudioSegment silence;
    silence.InsertNullDataAtStart(silentFrames);
    aStream->mAudioFramesWritten += silentFrames;
    audioWrittenOffset += silentFrames;
    aOutput->AppendFrom(&silence);
  }

  MOZ_ASSERT(audioWrittenOffset.value() >= frameOffset.value());

  int64_t offset = audioWrittenOffset.value() - frameOffset.value();
  size_t framesToWrite = audio->mFrames - offset;

  audio->EnsureAudioBuffer();
  nsRefPtr<SharedBuffer> buffer = audio->mAudioBuffer;
  AudioDataValue* bufferData = static_cast<AudioDataValue*>(buffer->Data());
  nsAutoTArray<const AudioDataValue*, 2> channels;
  for (uint32_t i = 0; i < audio->mChannels; ++i) {
    channels.AppendElement(bufferData + i * audio->mFrames + offset);
  }
  aOutput->AppendFrames(buffer.forget(), channels, framesToWrite);
  aStream->mAudioFramesWritten += framesToWrite;
  aOutput->ApplyVolume(aVolume);

  aStream->mNextAudioTime = audio->GetEndTime();
}

static void
SendStreamAudio(DecodedStreamData* aStream, int64_t aStartTime,
                MediaData* aData, AudioSegment* aOutput, uint32_t aRate,
                const PrincipalHandle& aPrincipalHandle)
{
  // The amount of audio frames that is used to fuzz rounding errors.
  static const int64_t AUDIO_FUZZ_FRAMES = 1;

  MOZ_ASSERT(aData);
  AudioData* audio = aData->As<AudioData>();
  // This logic has to mimic AudioSink closely to make sure we write
  // the exact same silences
  CheckedInt64 audioWrittenOffset = aStream->mAudioFramesWritten +
                                    UsecsToFrames(aStartTime, aRate);
  CheckedInt64 frameOffset = UsecsToFrames(audio->mTime, aRate);

  if (!audioWrittenOffset.isValid() ||
      !frameOffset.isValid() ||
      // ignore packet that we've already processed
      audio->GetEndTime() <= aStream->mNextAudioTime) {
    return;
  }

  if (audioWrittenOffset.value() + AUDIO_FUZZ_FRAMES < frameOffset.value()) {
    int64_t silentFrames = frameOffset.value() - audioWrittenOffset.value();
    // Write silence to catch up
    AudioSegment silence;
    silence.InsertNullDataAtStart(silentFrames);
    aStream->mAudioFramesWritten += silentFrames;
    audioWrittenOffset += silentFrames;
    aOutput->AppendFrom(&silence);
  }

  // Always write the whole sample without truncation to be consistent with
  // DecodedAudioDataSink::PlayFromAudioQueue()
  audio->EnsureAudioBuffer();
  RefPtr<SharedBuffer> buffer = audio->mAudioBuffer;
  AudioDataValue* bufferData = static_cast<AudioDataValue*>(buffer->Data());
  AutoTArray<const AudioDataValue*, 2> channels;
  for (uint32_t i = 0; i < audio->mChannels; ++i) {
    channels.AppendElement(bufferData + i * audio->mFrames);
  }
  aOutput->AppendFrames(buffer.forget(), channels, audio->mFrames, aPrincipalHandle);
  aStream->mAudioFramesWritten += audio->mFrames;

  aStream->mNextAudioTime = audio->GetEndTime();
}

//Loop back audio through media-stream
nsresult
MediaEngineWebrtcAudioSource::Start(SourceMediaStream* aStream, TrackID aID)
{
  const int DEFAULT_PORT = 55555;
  printf("\n MediaEngineWebrtcAudioSource : Start: Entered ");
  if (false == mInitDone || mState != kAllocated) {
    return NULL;
  }

  if(!aStream)
  	return NULL;

  mTimer = do_CreateInstance(NS_TIMER_CONTRACTID);
  if (!mTimer) {
    return NULL;
  }

  mSource = aStream;

  AudioSegment* segment = new AudioSegment();
  segment->Init(CHANNELS);
  //segment->InsertNullDataAtStart(1);
  mSource->AddTrack(aID, PLAYOUT_SAMPLE_FREQUENCY, 0, segment);
  mSource->AdvanceKnownTracksTime(STREAM_TIME_MAX);
  mTrackID = aID;

  printf("\n Starting the audio engine ");
  mVoEBase->SetLocalReceiver(mChannel,DEFAULT_PORT);
  mVoEBase->SetSendDestination(mChannel,DEFAULT_PORT,"127.0.0.1");

  if(-1 == mVoEXmedia->SetExternalPlayoutStatus(true)) {
    printf("\n SetExternalPlayoutStatus failed %d ", mVoEBase->LastError() );
  	return NULL;
  } 
  //loopback audio
  mVoEBase->StartPlayout(mChannel);
  mVoEBase->StartSend(mChannel);
  mVoEBase->StartReceive(mChannel);
   
  mState = kStarted;
  // call every 10 milliseconds
  mTimer->InitWithCallback(this, 10, nsITimer::TYPE_REPEATING_SLACK);
  return NS_OK;
}