WebRTC本地音频回调、选用音频采集设备及自定义输入音频

由于工作需要，开始研究WebRTC源码，现将如何把本地音频回调出来分享一下。

如果要使用Native WebRTC封装SDK，就要把RTC的一些基础能力暴露出来，本地视频、远端音视频都可以在相应的track里添加Sink就能拿到，但本地音频回调就没那么容易了，需要修改RTC的源码才可以。

首先来看一下AudioDeviceModule这个音频设备模块，它把录音和播放封装在一块，其中与录音相关的API有

//获取音频采集设备个数
virtual int16_t RecordingDevices() = 0;
//跟据设备索引获取设备名称和GUID(有的设备名称可能是一样，但GUID唯一)
virtual int32_t RecordingDeviceName(uint16_t index,
                                      char name[kAdmMaxDeviceNameSize],
                                      char guid[kAdmMaxGuidSize]) = 0;
//更换音频采集设备
virtual int32_t SetRecordingDevice(uint16_t index) = 0;
//初始化采集设备
virtual int32_t InitRecording() = 0;
//开始采集音频
virtual int32_t StartRecording() = 0;
//停止采集音频
virtual int32_t StopRecording() = 0;

这个模块会在创建webrtc::PeerConnectionFactoryInterface会用到

rtc::scoped_refptr<PeerConnectionFactoryInterface>
CreatePeerConnectionFactory(
    rtc::Thread* network_thread,
    rtc::Thread* worker_thread,
    rtc::Thread* signaling_thread,
    rtc::scoped_refptr<AudioDeviceModule> default_adm,
    rtc::scoped_refptr<AudioEncoderFactory> audio_encoder_factory,
    rtc::scoped_refptr<AudioDecoderFactory> audio_decoder_factory,
    std::unique_ptr<VideoEncoderFactory> video_encoder_factory,
    std::unique_ptr<VideoDecoderFactory> video_decoder_factory,
    rtc::scoped_refptr<AudioMixer> audio_mixer,
    rtc::scoped_refptr<AudioProcessing> audio_processing);

如果传入参数default_adm为nullptr，WebRTC会用AudioDeviceModuleImpl创建它，来看一下AudioDeviceModuleImpl的主要成员

//当前平台
PlatformType platform_type_ = kPlatformNotSupported;
//用于处理从audio_device_采集出来的音频，交由AudioTransport做一些合法性检查然后传输出去
AudioDeviceBuffer audio_device_buffer_;
//音频设备类，不同的平台有不同的实现 ，windows上用会AudioDeviceWindowsCore实现
std::unique_ptr<AudioDeviceGeneric> audio_device_;

audio_device_采集的音频会交由audio_device_buffer_处理，在由AudioTransport传输，来看一下AudioTransport

class AudioTransport {
 public:
  virtual int32_t RecordedDataIsAvailable(const void* audioSamples,
                                          const size_t nSamples,
                                          const size_t nBytesPerSample,
                                          const size_t nChannels,
                                          const uint32_t samplesPerSec,
                                          const uint32_t totalDelayMS,
                                          const int32_t clockDrift,
                                          const uint32_t currentMicLevel,
                                          const bool keyPressed,
                                          uint32_t& newMicLevel) = 0;  // NOLINT
// Implementation has to setup safe values for all specified out parameters.
 virtual int32_t NeedMorePlayData(const size_t nSamples,
                                   const size_t nBytesPerSample,
                                   const size_t nChannels,
                                   const uint32_t samplesPerSec,
                                   void* audioSamples,
                                   size_t& nSamplesOut,  // NOLINT
                                   int64_t* elapsed_time_ms,
                                   int64_t* ntp_time_ms) = 0;  // NOLINT
// Method to pull mixed render audio data from all active VoE channels.
// The data will not be passed as reference for audio processing internally.
virtual void PullRenderData(int bits_per_sample,
                              int sample_rate,
                              size_t number_of_channels,
                              size_t number_of_frames,
                              void* audio_data,
                              int64_t* elapsed_time_ms,
                              int64_t* ntp_time_ms) = 0;
 protected:
  virtual ~AudioTransport() {}
};

在WebRTC的默认实现中会用AudioTransportImpl实例化AudioTransport 。
以上分析可以看出如果想把设备控制权从WebRTC里拿出来可以自定义类继承至AudioDeviceModuleImpl，如果要把采集到的音频回调出来可以自定类继承至AudioTransportImpl。那么如何让音频回调是从我们自定义的类里创建的呢？再来看看默认的AudioTransportImpl是如何创建的。
通过源码可以看到在AudioState里有AudioTransportImpl对象，它是类对象，如果想让它从子类中创建必须为类对象指针，所以将它改为AudioTransportImpl*类型，将AudioState的构造改为

AudioState::AudioState(const AudioState::Config& config)
    : config_(config),audio_transport_(nullptr) {
  process_thread_checker_.Detach();
  RTC_DCHECK(config_.audio_mixer);
  RTC_DCHECK(config_.audio_device_module);
  audio_transport_ = audio_device_module()->CreateAudioTransPort(
      config_.audio_mixer, config_.audio_processing.get());
  needDelete_ = audio_transport_ == nullptr;
  if (needDelete_) {
    audio_transport_ = new AudioTransportImpl(config_.audio_mixer,
                                              config_.audio_processing.get());
  }
  RTC_DCHECK(audio_transport_);
}

上面的代码audio_device_module()是得到adm_对象，CreateAudioTransPort就是我们要增加的纯虚函数所以在AudioDeviceModule类中增加

 virtual AudioTransportImpl* CreateAudioTransPort(
      AudioMixer* mixer,
      AudioProcessing* audio_processing) = 0;

这个在AudioDeviceModuleImpl的实现为返回nullptr，这样如果没有子类实现它，结果就和原来的没有区别。

另外如果想要实现自定义输入，必须把WebRTC的采集停止掉，在AudioDeviceModule增加纯虚函数：

  virtual int32_t SendRecordedBuffer(const uint8_t* audio_data,
                                     uint32_t data_len,
                                     int bits_per_sample,
                                     int sample_rate,
                                     size_t number_of_channels) = 0;

在AudioDeviceModuleImpl实现：

int32_t AudioDeviceModuleImpl::SendRecordedBuffer(const uint8_t* audio_data,
                                                  uint32_t data_len,
                                                  int bits_per_sample,
                                                  int sample_rate,
                                                  size_t number_of_channels) {
  //if in external mode,shoudle control at subclass
  audio_device_buffer_.SetRecordingChannels(number_of_channels);
  audio_device_buffer_.SetRecordingSampleRate(sample_rate);
  audio_device_buffer_.SetRecordedBuffer(audio_data, (data_len >> 1) / number_of_channels);
  audio_device_buffer_.SetVQEData(0, 0);
  audio_device_buffer_.DeliverRecordedData();
  return 0;
}

整个的一个继承代码类似这样

//本地音频回调
class RecordedDataObserver {
public:
 virtual void onRecodedData(
  const void* audioSamples,
  const size_t nSamples,
  const size_t nBytesPerSample,
  const size_t nChannels, 
  const uint32_t samplesRate) = 0;
 virtual ~RecordedDataObserver() {}
};

namespace base {
  using namespace webrtc;
  class TaskQueueFactory;
  class AudioMixer;
  class AudioProcessing;
  class CustomizedAudioDeviceModule : public webrtc::AudioDeviceModuleImpl{
  public:
  	static rtc::scoped_refptr<webrtc::AudioDeviceModule> CreateADM();
   CustomizedAudioDeviceModule(AudioLayer audio_layer,
    webrtc::TaskQueueFactory* task_queue_factory);
   virtual ~CustomizedAudioDeviceModule();
   class AudioRecordCallBack : public AudioTransportImpl {
   public:
    AudioRecordCallBack(webrtc::AudioMixer* mixer, webrtc::AudioProcessing* audio_processing);
    virtual ~AudioRecordCallBack();
    int32_t RecordedDataIsAvailable(const void* audioSamples,
     const size_t nSamples,
     const size_t nBytesPerSample,
     const size_t nChannels,
     const uint32_t samplesPerSec,
     const uint32_t totalDelayMS,
     const int32_t clockDrift,
     const uint32_t currentMicLevel,
     const bool keyPressed,
     uint32_t& newMicLevel) override;
    void RegisterObserver(RecordedDataObserver* dataObserver);
   private:
    RecordedDataObserver* recorderDataOb;
   };
   AudioTransportImpl* CreateAudioTransPort(
    webrtc::AudioMixer* mixer,
    webrtc::AudioProcessing* audio_processing) override;
   bool IsExternalAudioInput() const override;
   void SetExternalAudioMode(bool bExternal);
   void RegisterObserver(RecordedDataObserver* dataObserver);
  private:
   std::unique_ptr<AudioRecordCallBack> audioRecordCb_;
   bool externalAudio_;
   RecordedDataObserver* dataObserver_;
   };
}

实现文件

rtc::scoped_refptr<webrtc::AudioDeviceModule> base::CustomizedAudioDeviceModule::CreateADM()
{
 // Create the generic reference counted (platform independent) implementation.
 rtc::scoped_refptr<base::CustomizedAudioDeviceModule> audioDevice(
      new rtc::RefCountedObject<base::CustomizedAudioDeviceModule>(webrtc::AudioDeviceModule::kPlatformDefaultAudio,
    &webrtc::GlobalTaskQueueFactory()));
    // Ensure that the current platform is supported.
 if (audioDevice->CheckPlatform() == -1) {
  return nullptr;
 }
 // Ensure that the generic audio buffer can communicate with the platform
 // specific parts.
 if (audioDevice->AttachAudioBuffer() == -1) {
  return nullptr;
 }
 return audioDevice;
}

base::CustomizedAudioDeviceModule::~CustomizedAudioDeviceModule() {
}
webrtc::AudioTransportImpl* base::CustomizedAudioDeviceModule::CreateAudioTransPort(
 webrtc::AudioMixer* mixer, 
 webrtc::AudioProcessing* audio_processing) {

audioRecordCb_ = std::make_unique<AudioRecordCallBack>(mixer, audio_processing);
 if (dataObserver_) {
  audioRecordCb_->RegisterObserver(dataObserver_);
 }
 return audioRecordCb_.get();

 }

bool base::CustomizedAudioDeviceModule::IsExternalAudioInput() const {
	return externalAudio_;
}
void base::CustomizedAudioDeviceModule::SetExternalAudioMode(bool bExternal) {
	externalAudio_ = bExternal;
}

void base::CustomizedAudioDeviceModule::RegisterObserver(RecordedDataObserver* dataObserver) {
//要在CreatePeerConnectionFactory后调用
 RTC_CHECK(dataObserver);
 RTC_CHECK(audioRecordCb_);
 dataObserver_ = dataObserver;
 audioRecordCb_->RegisterObserver(dataObserver);
}

base::CustomizedAudioDeviceModule::AudioRecordCallBack::AudioRecordCallBack(
 webrtc::AudioMixer* mixer,
 webrtc::AudioProcessing* audio_processing) : 
 AudioTransportImpl(mixer,audio_processing), 
 recorderDataOb(nullptr) {
 }

base::CustomizedAudioDeviceModule::AudioRecordCallBack::~AudioRecordCallBack() {
}

int32_t base::CustomizedAudioDeviceModule::AudioRecordCallBack::RecordedDataIsAvailable( const void* audioSamples, 
 const size_t nSamples, 
 const size_t nBytesPerSample, 
 const size_t nChannels, 
 const uint32_t samplesPerSec, 
 const uint32_t totalDelayMS, 
 const int32_t clockDrift, 
 const uint32_t currentMicLevel, 
 const bool keyPressed, 
 uint32_t& newMicLevel) {
	
	if (recorderDataOb) {
  recorderDataOb->onRecodedData(audioSamples, nSamples, nBytesPerSample, nChannels,samplesPerSec);
 }
 
 return AudioTransportImpl::RecordedDataIsAvailable(
  audioSamples, 
  nSamples, 
  nBytesPerSample, 
  nChannels, 
  samplesPerSec, 
  totalDelayMS, 
  clockDrift, 
  currentMicLevel, 
  keyPressed, 
  newMicLevel);
 }
void base::CustomizedAudioDeviceModule::AudioRecordCallBack::RegisterObserver(RecordedDataObserver* dataObserver) {
recorderDataOb = dataObserver;
}

在webrtc::CreatePeerConnectionFactory的时候把adm对象传进去，需要注意的是，adm对象要由worker_thread创建，原因是AudioDeviceModuleImpl中的audio_device_buffer_要在worker_thread中运行，所以adm的StartRecording()、StopRecording()，也要在worker_thread中运行。
adm对象销毁也要由worker_thread销毁。

换音频采集设置调用adm对象的SetRecordingDevice(uint16_t index)，需要StopRecording,SetRecordingDevice,InitRecording,StartRecording。

自定义输入音频要增加接口类似这种：

int32_t AudioDeviceModuleImpl::SendRecordedBuffer(const uint8_t* audio_data,
                                                  uint32_t data_len,
                                                  int bits_per_sample,
                                                  int sample_rate,
                                                  size_t number_of_channels) {
  if (!external_audio_input_)
    return -1;
  audio_device_buffer_.SetRecordingChannels(number_of_channels);
  audio_device_buffer_.SetRecordingSampleRate(sample_rate);
  audio_device_buffer_.SetRecordedBuffer(audio_data, (data_len >> 1) / number_of_channels);
  audio_device_buffer_.SetVQEData(0, 0);
  audio_device_buffer_.DeliverRecordedData();
  return 0;
}

本人刚开始学习WebRTC，难免有不当之处，欢迎指正。
如果有更好的实现，欢迎分享。