WebRtc VoiceEngine代码解析

WebRtc中VoiceEngine可以完成大部分的VOIP相关人物，包括采集、自动增益、噪声消除、回声抑制、编解码、ＲＴＰ传输。下边我们通过代码来解析Voe中处理流程；

创建VoiceEngine和VoEBase

[cpp]
VoiceEngine* _vePtr = VoiceEngine::Create();                               //创建VoiceEngine  
VoEBase* _veBasePtr = VoEBase::GetInterface(_vePtr);            //创建VoeBase  所有Voe相关操作通过这个共有类  
_veBasePtr->Init()                                                                                 //创建整个Voe处理线程

VoiceEngine* _vePtr = VoiceEngine::Create();                              //创建VoiceEngine
VoEBase* _veBasePtr = VoEBase::GetInterface(_vePtr);            //创建VoeBase  所有Voe相关操作通过这个共有类
_veBasePtr->Init()                                                                               //创建整个Voe处理线程
重点就在_veBasePtr->Init()  它会创建voe线程，线程负责采集、数字信号处理、编码、rtp传输。

[cpp]
int VoEBaseImpl::Init(AudioDeviceModule* external_adm,   AudioProcessing* audioproc) 
{ 
    _shared->process_thread()；   //创建voe线程  
    _shared->process_thread()->Start()； 
    _shared->audio_device()->Init()； 
    
 
}

int VoEBaseImpl::Init(AudioDeviceModule* external_adm,   AudioProcessing* audioproc)
{
    _shared->process_thread()；   //创建voe线程
    _shared->process_thread()->Start()；
    _shared->audio_device()->Init()；

}audio_device()->Init()重载了int32_t AudioDeviceWindowsWave::Init()（windowns平台），别的平台是别的函数，基本差不多，在这个Init中，创建了ThreadProcess线程，ThreadProcess线程负责所有的音频流程，从设备获取音频数据包。

[cpp]
bool AudioDeviceWindowsWave::ThreadProcess() 
{ 
   while ((nRecordedBytes = RecProc(recTime)) > 0)； 
}

bool AudioDeviceWindowsWave::ThreadProcess()
{
   while ((nRecordedBytes = RecProc(recTime)) > 0)；
}处理过程在RecProc

[cpp]
int32_t AudioDeviceWindowsWave::RecProc(LONGLONG& consumedTime) 
{ 
     _ptrAudioBuffer->DeliverRecordedData(); <SPAN style="FONT-FAMILY: Arial, Helvetica, sans-serif">}</SPAN>

int32_t AudioDeviceWindowsWave::RecProc(LONGLONG& consumedTime)
{
     _ptrAudioBuffer->DeliverRecordedData(); }

[cpp]
int32_t AudioDeviceBuffer::DeliverRecordedData() 
{ 
     _ptrCbAudioTransport->RecordedDataIsAvailable(); 
}

int32_t AudioDeviceBuffer::DeliverRecordedData()
{
     _ptrCbAudioTransport->RecordedDataIsAvailable();
}

RecordedDataIsAvailable是虚函数，被VoeBase重载

[cpp]
int32_t VoEBaseImpl::RecordedDataIsAvailable( 
        const void* audioSamples, 
        uint32_t nSamples, 
        uint8_t nBytesPerSample, 
        uint8_t nChannels, 
        uint32_t samplesPerSec, 
        uint32_t totalDelayMS, 
        int32_t clockDrift, 
        uint32_t currentMicLevel, 
        bool keyPressed, 
        uint32_t& newMicLevel) 
{ 
     _shared->transmit_mixer()->DemuxAndMix(); 
    _shared->transmit_mixer()->EncodeAndSend(); 
}

int32_t VoEBaseImpl::RecordedDataIsAvailable(
        const void* audioSamples,
        uint32_t nSamples,
        uint8_t nBytesPerSample,
        uint8_t nChannels,
        uint32_t samplesPerSec,
        uint32_t totalDelayMS,
        int32_t clockDrift,
        uint32_t currentMicLevel,
        bool keyPressed,
        uint32_t& newMicLevel)
{
     _shared->transmit_mixer()->DemuxAndMix();
    _shared->transmit_mixer()->EncodeAndSend();
}
DemuxAndMix() 从字面意思是分路与混合，这个函数，主要负责AudioProcess的所有过程，包括Aec,Aecm,AGC,DTMF，遍历所有channel;

[cpp]
TransmitMixer::DemuxAndMix() 
{ 
    Channel* channelPtr = sc.GetFirstChannel(iterator); 
    while (channelPtr != NULL) 
    { 
        if (channelPtr->InputIsOnHold()) 
        { 
            channelPtr->UpdateLocalTimeStamp(); 
        } else if (channelPtr->Sending()) 
        { 
            // Demultiplex makes a copy of its input.  
            channelPtr->Demultiplex(_audioFrame); 
            channelPtr->PrepareEncodeAndSend(_audioFrame.sample_rate_hz_); 
        } 
        channelPtr = sc.GetNextChannel(iterator); 
    } 
 
}

TransmitMixer::DemuxAndMix()
{
    Channel* channelPtr = sc.GetFirstChannel(iterator);
    while (channelPtr != NULL)
    {
        if (channelPtr->InputIsOnHold())
        {
            channelPtr->UpdateLocalTimeStamp();
        } else if (channelPtr->Sending())
        {
            // Demultiplex makes a copy of its input.
            channelPtr->Demultiplex(_audioFrame);
            channelPtr->PrepareEncodeAndSend(_audioFrame.sample_rate_hz_);
        }
        channelPtr = sc.GetNextChannel(iterator);
    }

}
Channel::Demutiplex(),基本上没有什么具体任务，就是把audioFrame里边的数据拷贝到channel自身, webrtc是client解决方案，对于client只认为有一个audio source,但可以有多个channel，每个channel中都有audio process，所以需要把数据copy到每个channel.

只有就是数据处理 PrepareEncodeAndSend()

[cpp]
Channel::PrepareEncodeAndSend(int mixingFrequency) 
{ 
   if (_inputFilePlaying) 
    { 
        MixOrReplaceAudioWithFile(mixingFrequency); //如果使用了voeFile::PlayFileAsMic()；则从文件读取10ms数据，并覆盖audio buffer  
    } 
 
    if (_mute) 
    { 
        AudioFrameOperations::Mute(_audioFrame);//当然如果设置mutex，则memset 0   
    } 
   if (_inputExternalMedia) 
   { 
   _inputExternalMediaCallbackPtr->Process();  //所过设置了ExternalMedia,自己的audio处理过程，就是在这里调用的  
   } 
  InsertInbandDtmfTone();                     //添加DTMF音频  
 _rtpAudioProc->ProcessStream(&_audioFrame);  // 真正的GIPS牛逼代码，audio process过程: Aec Aecm AGC   
}

Channel::PrepareEncodeAndSend(int mixingFrequency)
{
   if (_inputFilePlaying)
    {
        MixOrReplaceAudioWithFile(mixingFrequency); //如果使用了voeFile::PlayFileAsMic()；则从文件读取10ms数据，并覆盖audio buffer
    }

if (_mute)
    {
        AudioFrameOperations::Mute(_audioFrame);//当然如果设置mutex，则memset 0
    }
   if (_inputExternalMedia)
   {
   _inputExternalMediaCallbackPtr->Process();  //所过设置了ExternalMedia,自己的audio处理过程，就是在这里调用的
   }
  InsertInbandDtmfTone();                     //添加DTMF音频
 _rtpAudioProc->ProcessStream(&_audioFrame);  // 真正的GIPS牛逼代码，audio process过程: Aec Aecm AGC
}
int AudioProcessingImpl::ProcessStream(AudioFrame* frame) 就是上述调用的_rtpAudioProc->ProcessStream();

以上是DemuxAndMix（）过程，之后就是EncodeAndSend()过程，至此整个voe数据处理流程分析结束；

关于Audio Process则是另外一个大话题；

总结一下几点：

1.  VoeBase提供大部分的对外接口

2. Channel：继承了大部分的音频功能；