基于 RTC 場景下要解決的聲音的問題，WebRTC 有一個大體如下圖所示的音頻處理流水線：

WebRTC 的音頻處理流水線，不是一次性建立起來的，而是分階段分步驟建立的。整體而言，可以認為這個流水線分兩個階段建立，或者可以認為這個流水線分為兩部分：一部分可稱為靜態流水線，另一部分可稱為動態流水線，或者也可以稱為前端和后端。靜態流水線，在某個時間點建立一次，隨后在整個 WebRTC 通信過程中基本保持不變；動態流水線則在通信過程中，可能出現較為頻繁的變動，如本地打開或禁用錄制啟動發送或停止發送音頻數據，遠端發送者加入或退出頻道等，都會改變動態流水線。

如此，WebRTC 的音頻處理流水線大致如下圖所示：

WebRTC 音頻的靜態流水線，建立之后，其相關節點狀態由 AudioState 維護和管理。WebRTC 音頻的靜態流水線，主要包括 AudioDeviceModule，AudioProcessing，和 AudioMixer 等，其中 AudioDeviceModule 用于采集和播放音頻數據，AudioProcessing 主要用于對錄制的音頻數據做初始處理，如回聲消除，降噪等，AudioMixer 主要用于對遠端發送過來的音頻數據做混音。

WebRTC 音頻的靜態流水線在 WebRtcVoiceEngine 初始化時建立：

void WebRtcVoiceEngine::Init() {RTC_DCHECK(worker_thread_checker_.IsCurrent());RTC_LOG(LS_INFO) << "WebRtcVoiceEngine::Init";// TaskQueue expects to be created/destroyed on the same thread.low_priority_worker_queue_.reset(new rtc::TaskQueue(task_queue_factory_->CreateTaskQueue("rtc-low-prio", webrtc::TaskQueueFactory::Priority::LOW)));// Load our audio codec lists.RTC_LOG(LS_INFO) << "Supported send codecs in order of preference:";send_codecs_ = CollectCodecs(encoder_factory_->GetSupportedEncoders());for (const AudioCodec& codec : send_codecs_) {RTC_LOG(LS_INFO) << ToString(codec);}RTC_LOG(LS_INFO) << "Supported recv codecs in order of preference:";recv_codecs_ = CollectCodecs(decoder_factory_->GetSupportedDecoders());for (const AudioCodec& codec : recv_codecs_) {RTC_LOG(LS_INFO) << ToString(codec);}#if defined(WEBRTC_INCLUDE_INTERNAL_AUDIO_DEVICE)// No ADM supplied? Create a default one.if (!adm_) {adm_ = webrtc::AudioDeviceModule::Create(webrtc::AudioDeviceModule::kPlatformDefaultAudio, task_queue_factory_);} #endif // WEBRTC_INCLUDE_INTERNAL_AUDIO_DEVICERTC_CHECK(adm());webrtc::adm_helpers::Init(adm());webrtc::apm_helpers::Init(apm());// Set up AudioState.{webrtc::AudioState::Config config;if (audio_mixer_) {config.audio_mixer = audio_mixer_;} else {config.audio_mixer = webrtc::AudioMixerImpl::Create();}config.audio_processing = apm_;config.audio_device_module = adm_;audio_state_ = webrtc::AudioState::Create(config);}// Connect the ADM to our audio path.adm()->RegisterAudioCallback(audio_state()->audio_transport());// Set default engine options.{AudioOptions options;options.echo_cancellation = true;options.auto_gain_control = true;options.noise_suppression = true;options.highpass_filter = true;options.stereo_swapping = false;options.audio_jitter_buffer_max_packets = 200;options.audio_jitter_buffer_fast_accelerate = false;options.audio_jitter_buffer_min_delay_ms = 0;options.audio_jitter_buffer_enable_rtx_handling = false;options.typing_detection = true;options.experimental_agc = false;options.extended_filter_aec = false;options.delay_agnostic_aec = false;options.experimental_ns = false;options.residual_echo_detector = true;bool error = ApplyOptions(options);RTC_DCHECK(error);}initialized_ = true; }

WebRTC 音頻靜態流水線的幾個節點，由外部創建好注入進來，并通過 AudioState 創建的 AudioTransport 連接起來：WebRtcVoiceEngine::Init()
根據需要創建 AudioDeviceModule，然后創建 AudioState，AudioState 用注入的 AudioMixer 和 AudioProcessing 創建 AudioTransport，WebRtcVoiceEngine::Init() 隨后將 AudioState 創建的 AudioTransport 作為 AudioCallback 注冊給 AudioDeviceModule。WebRTC 音頻靜態流水線在 WebRtcVoiceEngine::Init() 中建立。動態流水線和靜態流水線在 AudioState 和 AudioTransport 處交匯。

AudioTransport

AudioDeviceModule 在錄到音頻數據時，將音頻數據送給 AudioTransport，在播放音頻時，也向 AudioTransport 拿要播放的音頻數據。AudioTransport 是錄制的音頻數據和播放的音頻數據的交匯點。

AudioTransport 接口的定義 (位于 webrtc/src/modules/audio_device/include/audio_device_defines.h) 如下：

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, // NOLINTint64_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() {} };

WebRTC 中的 AudioTransport 接口實現 AudioTransportImpl 定義 (位于 webrtc/src/audio/audio_transport_impl.h) 如下：

namespace webrtc {class AudioSendStream;class AudioTransportImpl : public AudioTransport {public:AudioTransportImpl(AudioMixer* mixer, AudioProcessing* audio_processing);~AudioTransportImpl() override;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;int32_t NeedMorePlayData(const size_t nSamples,const size_t nBytesPerSample,const size_t nChannels,const uint32_t samplesPerSec,void* audioSamples,size_t& nSamplesOut,int64_t* elapsed_time_ms,int64_t* ntp_time_ms) override;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) override;void UpdateSendingStreams(std::vector<AudioSendStream*> streams,int send_sample_rate_hz,size_t send_num_channels);void SetStereoChannelSwapping(bool enable);bool typing_noise_detected() const;const voe::AudioLevel& audio_level() const { return audio_level_; }private:// Shared.AudioProcessing* audio_processing_ = nullptr;// Capture side.rtc::CriticalSection capture_lock_;std::vector<AudioSendStream*> sending_streams_ RTC_GUARDED_BY(capture_lock_);int send_sample_rate_hz_ RTC_GUARDED_BY(capture_lock_) = 8000;size_t send_num_channels_ RTC_GUARDED_BY(capture_lock_) = 1;bool typing_noise_detected_ RTC_GUARDED_BY(capture_lock_) = false;bool swap_stereo_channels_ RTC_GUARDED_BY(capture_lock_) = false;PushResampler<int16_t> capture_resampler_;voe::AudioLevel audio_level_;TypingDetection typing_detection_;// Render side.rtc::scoped_refptr<AudioMixer> mixer_;AudioFrame mixed_frame_;// Converts mixed audio to the audio device output rate.PushResampler<int16_t> render_resampler_;RTC_DISALLOW_IMPLICIT_CONSTRUCTORS(AudioTransportImpl); }; } // namespace webrtc

AudioTransportImpl 實現的 AudioTransport 接口的要求的操作：將錄制的音頻數據送出去，及獲取播放的音頻數據。

AudioTransportImpl 對象創建時保存注入的 AudioMixer 和 AudioProcessing 對象指針：

AudioTransportImpl::AudioTransportImpl(AudioMixer* mixer,AudioProcessing* audio_processing): audio_processing_(audio_processing), mixer_(mixer) {RTC_DCHECK(mixer);RTC_DCHECK(audio_processing); }AudioTransportImpl::~AudioTransportImpl() {}

播放

AudioDeviceModule 可以通過 AudioTransportImpl 的 NeedMorePlayData() 和 PullRenderData() 獲取用于播放的音頻數據：

// Resample audio in |frame| to given sample rate preserving the // channel count and place the result in |destination|. int Resample(const AudioFrame& frame,const int destination_sample_rate,PushResampler<int16_t>* resampler,int16_t* destination) {const int number_of_channels = static_cast<int>(frame.num_channels_);const int target_number_of_samples_per_channel =destination_sample_rate / 100;resampler->InitializeIfNeeded(frame.sample_rate_hz_, destination_sample_rate,number_of_channels);// TODO(yujo): make resampler take an AudioFrame, and add special case// handling of muted frames.return resampler->Resample(frame.data(), frame.samples_per_channel_ * number_of_channels,destination, number_of_channels * target_number_of_samples_per_channel); } . . . . . . // Mix all received streams, feed the result to the AudioProcessing module, then // resample the result to the requested output rate. int32_t AudioTransportImpl::NeedMorePlayData(const size_t nSamples,const size_t nBytesPerSample,const size_t nChannels,const uint32_t samplesPerSec,void* audioSamples,size_t& nSamplesOut,int64_t* elapsed_time_ms,int64_t* ntp_time_ms) {RTC_DCHECK_EQ(sizeof(int16_t) * nChannels, nBytesPerSample);RTC_DCHECK_GE(nChannels, 1);RTC_DCHECK_LE(nChannels, 2);RTC_DCHECK_GE(samplesPerSec,static_cast<uint32_t>(AudioProcessing::NativeRate::kSampleRate8kHz));// 100 = 1 second / data duration (10 ms).RTC_DCHECK_EQ(nSamples * 100, samplesPerSec);RTC_DCHECK_LE(nBytesPerSample * nSamples * nChannels,AudioFrame::kMaxDataSizeBytes);mixer_->Mix(nChannels, &mixed_frame_);*elapsed_time_ms = mixed_frame_.elapsed_time_ms_;*ntp_time_ms = mixed_frame_.ntp_time_ms_;const auto error = audio_processing_->ProcessReverseStream(&mixed_frame_);RTC_DCHECK_EQ(error, AudioProcessing::kNoError);nSamplesOut = Resample(mixed_frame_, samplesPerSec, &render_resampler_,static_cast<int16_t*>(audioSamples));RTC_DCHECK_EQ(nSamplesOut, nChannels * nSamples);return 0; }// Used by Chromium - same as NeedMorePlayData() but because Chrome has its // own APM instance, does not call audio_processing_->ProcessReverseStream(). void AudioTransportImpl::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) {RTC_DCHECK_EQ(bits_per_sample, 16);RTC_DCHECK_GE(number_of_channels, 1);RTC_DCHECK_GE(sample_rate, AudioProcessing::NativeRate::kSampleRate8kHz);// 100 = 1 second / data duration (10 ms).RTC_DCHECK_EQ(number_of_frames * 100, sample_rate);// 8 = bits per byte.RTC_DCHECK_LE(bits_per_sample / 8 * number_of_frames * number_of_channels,AudioFrame::kMaxDataSizeBytes);mixer_->Mix(number_of_channels, &mixed_frame_);*elapsed_time_ms = mixed_frame_.elapsed_time_ms_;*ntp_time_ms = mixed_frame_.ntp_time_ms_;auto output_samples = Resample(mixed_frame_, sample_rate, &render_resampler_,static_cast<int16_t*>(audio_data));RTC_DCHECK_EQ(output_samples, number_of_channels * number_of_frames); }

NeedMorePlayData() 執行如下操作：

通過 AudioMixer 對遠端的音頻流做混音。AudioTransportImpl 沒有直接與 AudioReceiveStream 建立連接，但 AudioMixer 的 source 實際上是 AudioReceiveStream。

把混音后的音頻數據送進 AudioProcessing。主要用于回聲消除。

對混音后的音頻數據重采樣。如我們前面看到的 AudioMixerImpl 的實現，它的混音輸出采樣率有它自己的決定邏輯，可能與 AudioDeviceModule 設備需要的不一致，這里需要按照設備的需要，將混音后的數據重采樣到設備需要的采樣率。

返回重采樣的數據。

PullRenderData() 的操作與 NeedMorePlayData() 僅有的區別在于，沒有上面的第 2 步。如注釋中所說，在某些情況下，混音之后的數據無需送進 AudioProcessing。此外，PullRenderData() 要求調用者只能請求單通道的音頻數據，NeedMorePlayData() 的調用者則可以請求單通道或立體聲的。

重采樣和混音的更多細節，這里不再贅述。

錄制

AudioDeviceModule 可以通過 AudioTransport 的 RecordedDataIsAvailable() 把錄制的數據送出去：

// We want to process at the lowest sample rate and channel count possible // without losing information. Choose the lowest native rate at least equal to // the minimum of input and codec rates, choose lowest channel count, and // configure the audio frame. void InitializeCaptureFrame(int input_sample_rate,int send_sample_rate_hz,size_t input_num_channels,size_t send_num_channels,AudioFrame* audio_frame) {RTC_DCHECK(audio_frame);int min_processing_rate_hz = std::min(input_sample_rate, send_sample_rate_hz);for (int native_rate_hz : AudioProcessing::kNativeSampleRatesHz) {audio_frame->sample_rate_hz_ = native_rate_hz;if (audio_frame->sample_rate_hz_ >= min_processing_rate_hz) {break;}}audio_frame->num_channels_ = std::min(input_num_channels, send_num_channels); }void ProcessCaptureFrame(uint32_t delay_ms,bool key_pressed,bool swap_stereo_channels,AudioProcessing* audio_processing,AudioFrame* audio_frame) {RTC_DCHECK(audio_processing);RTC_DCHECK(audio_frame);audio_processing->set_stream_delay_ms(delay_ms);audio_processing->set_stream_key_pressed(key_pressed);int error = audio_processing->ProcessStream(audio_frame);RTC_DCHECK_EQ(0, error) << "ProcessStream() error: " << error;if (swap_stereo_channels) {AudioFrameOperations::SwapStereoChannels(audio_frame);} } . . . . . . // Not used in Chromium. Process captured audio and distribute to all sending // streams, and try to do this at the lowest possible sample rate. int32_t AudioTransportImpl::RecordedDataIsAvailable(const void* audio_data,const size_t number_of_frames,const size_t bytes_per_sample,const size_t number_of_channels,const uint32_t sample_rate,const uint32_t audio_delay_milliseconds,const int32_t /*clock_drift*/,const uint32_t /*volume*/,const bool key_pressed,uint32_t& /*new_mic_volume*/) { // NOLINT: to avoid changing APIsRTC_DCHECK(audio_data);RTC_DCHECK_GE(number_of_channels, 1);RTC_DCHECK_LE(number_of_channels, 2);RTC_DCHECK_EQ(2 * number_of_channels, bytes_per_sample);RTC_DCHECK_GE(sample_rate, AudioProcessing::NativeRate::kSampleRate8kHz);// 100 = 1 second / data duration (10 ms).RTC_DCHECK_EQ(number_of_frames * 100, sample_rate);RTC_DCHECK_LE(bytes_per_sample * number_of_frames * number_of_channels,AudioFrame::kMaxDataSizeBytes);int send_sample_rate_hz = 0;size_t send_num_channels = 0;bool swap_stereo_channels = false;{rtc::CritScope lock(&capture_lock_);send_sample_rate_hz = send_sample_rate_hz_;send_num_channels = send_num_channels_;swap_stereo_channels = swap_stereo_channels_;}std::unique_ptr<AudioFrame> audio_frame(new AudioFrame());InitializeCaptureFrame(sample_rate, send_sample_rate_hz, number_of_channels,send_num_channels, audio_frame.get());voe::RemixAndResample(static_cast<const int16_t*>(audio_data),number_of_frames, number_of_channels, sample_rate,&capture_resampler_, audio_frame.get());ProcessCaptureFrame(audio_delay_milliseconds, key_pressed,swap_stereo_channels, audio_processing_,audio_frame.get());// Typing detection (utilizes the APM/VAD decision). We let the VAD determine// if we're using this feature or not.// TODO(solenberg): GetConfig() takes a lock. Work around that.bool typing_detected = false;if (audio_processing_->GetConfig().voice_detection.enabled) {if (audio_frame->vad_activity_ != AudioFrame::kVadUnknown) {bool vad_active = audio_frame->vad_activity_ == AudioFrame::kVadActive;typing_detected = typing_detection_.Process(key_pressed, vad_active);}}// Measure audio level of speech after all processing.double sample_duration = static_cast<double>(number_of_frames) / sample_rate;audio_level_.ComputeLevel(*audio_frame, sample_duration);// Copy frame and push to each sending stream. The copy is required since an// encoding task will be posted internally to each stream.{rtc::CritScope lock(&capture_lock_);typing_noise_detected_ = typing_detected;RTC_DCHECK_GT(audio_frame->samples_per_channel_, 0);if (!sending_streams_.empty()) {auto it = sending_streams_.begin();while (++it != sending_streams_.end()) {std::unique_ptr<AudioFrame> audio_frame_copy(new AudioFrame());audio_frame_copy->CopyFrom(*audio_frame);(*it)->SendAudioData(std::move(audio_frame_copy));}// Send the original frame to the first stream w/o copying.(*sending_streams_.begin())->SendAudioData(std::move(audio_frame));}}return 0; } . . . . . . void AudioTransportImpl::UpdateSendingStreams(std::vector<AudioSendStream*> streams,int send_sample_rate_hz,size_t send_num_channels) {rtc::CritScope lock(&capture_lock_);sending_streams_ = std::move(streams);send_sample_rate_hz_ = send_sample_rate_hz;send_num_channels_ = send_num_channels; }

AudioTransportImpl 的 RecordedDataIsAvailable() 拿到錄音得到的音頻數據，對音頻數據執行如下操作：

重采樣。AudioDeviceModule 送進來的音頻數據的采樣率是設備采集的原始采樣率，這里需要把采集的音頻數據轉換為適合 WebRTC 的音頻處理模塊如降噪和回聲消除等模塊及編碼器處理的采樣率。

對音頻數據做回聲消除，降噪等處理。

把音頻數據送出去，給注入的 AudioSendStream。

AudioState

AudioState 既是 WebRTC 音頻流水線節點的容器，也可用于對這些節點做一些管理和控制，以及將不同節點粘起來構成流水線。AudioState 接口的定義 (位于 webrtc/src/call/audio_state.h) 如下：

namespace webrtc {class AudioTransport;// AudioState holds the state which must be shared between multiple instances of // webrtc::Call for audio processing purposes. class AudioState : public rtc::RefCountInterface {public:struct Config {Config();~Config();// The audio mixer connected to active receive streams. One per// AudioState.rtc::scoped_refptr<AudioMixer> audio_mixer;// The audio processing module.rtc::scoped_refptr<webrtc::AudioProcessing> audio_processing;// TODO(solenberg): Temporary: audio device module.rtc::scoped_refptr<webrtc::AudioDeviceModule> audio_device_module;};struct Stats {// Audio peak level (max(abs())), linearly on the interval [0,32767].int32_t audio_level = -1;// See:// https://w3c.github.io/webrtc-stats/#dom-rtcmediastreamtrackstats-totalaudioenergydouble total_energy = 0.0f;double total_duration = 0.0f;};virtual AudioProcessing* audio_processing() = 0;virtual AudioTransport* audio_transport() = 0;// Enable/disable playout of the audio channels. Enabled by default.// This will stop playout of the underlying audio device but start a task// which will poll for audio data every 10ms to ensure that audio processing// happens and the audio stats are updated.virtual void SetPlayout(bool enabled) = 0;// Enable/disable recording of the audio channels. Enabled by default.// This will stop recording of the underlying audio device and no audio// packets will be encoded or transmitted.virtual void SetRecording(bool enabled) = 0;virtual Stats GetAudioInputStats() const = 0;virtual void SetStereoChannelSwapping(bool enable) = 0;static rtc::scoped_refptr<AudioState> Create(const AudioState::Config& config);~AudioState() override {} }; } // namespace webrtc

作為音頻處理流水線的容器，AudioState 接口主要包含了 AudioMixer、AudioProcessing 和 AudioDeviceModule。AudioState 對象的創建主要通過靜態工廠方法完成。

WebRTC 提供了 AudioState 接口的實現，該實現的定義 (位于 webrtc/src/audio/audio_state.h ) 如下：

namespace webrtc {class AudioSendStream; class AudioReceiveStream;namespace internal {class AudioState : public webrtc::AudioState {public:explicit AudioState(const AudioState::Config& config);~AudioState() override;AudioProcessing* audio_processing() override;AudioTransport* audio_transport() override;void SetPlayout(bool enabled) override;void SetRecording(bool enabled) override;Stats GetAudioInputStats() const override;void SetStereoChannelSwapping(bool enable) override;AudioDeviceModule* audio_device_module() {RTC_DCHECK(config_.audio_device_module);return config_.audio_device_module.get();}bool typing_noise_detected() const;void AddReceivingStream(webrtc::AudioReceiveStream* stream);void RemoveReceivingStream(webrtc::AudioReceiveStream* stream);void AddSendingStream(webrtc::AudioSendStream* stream,int sample_rate_hz,size_t num_channels);void RemoveSendingStream(webrtc::AudioSendStream* stream);private:void UpdateAudioTransportWithSendingStreams();rtc::ThreadChecker thread_checker_;rtc::ThreadChecker process_thread_checker_;const webrtc::AudioState::Config config_;bool recording_enabled_ = true;bool playout_enabled_ = true;// Transports mixed audio from the mixer to the audio device and// recorded audio to the sending streams.AudioTransportImpl audio_transport_;// Null audio poller is used to continue polling the audio streams if audio// playout is disabled so that audio processing still happens and the audio// stats are still updated.std::unique_ptr<NullAudioPoller> null_audio_poller_;std::unordered_set<webrtc::AudioReceiveStream*> receiving_streams_;struct StreamProperties {int sample_rate_hz = 0;size_t num_channels = 0;};std::map<webrtc::AudioSendStream*, StreamProperties> sending_streams_;RTC_DISALLOW_IMPLICIT_CONSTRUCTORS(AudioState); }; } // namespace internal

如上面看到的，AudioState 對象在 WebRtcVoiceEngine::Init() 創建。AudioState 的 Create() 函數定義 (位于 webrtc/src/audio/audio_state.cpp) 如下：

rtc::scoped_refptr<AudioState> AudioState::Create(const AudioState::Config& config) {return new rtc::RefCountedObject<internal::AudioState>(config); }

WebRTC 中實際在用的 AudioState 對象為 webrtc::internal::AudioState 的對象。

AudioState 完成的工作主要為：

音頻流水線的節點的 Getter

管理音頻靜態流水線和動態流水線的連接，也就是添加和移除 AudioSendStream 和 AudioReceiveStream：把 AudioSendStream 注入給 AudioTransportImpl 或從中移除，把 AudioReceiveStream 添加到 AudioMixer 或從中移除。這些接口主要由 webrtc::Call 使用。

控制播放和錄制的啟動和停止。

webrtc::internal::AudioState 的具體實現 (位于 webrtc/src/audio/audio_state.cpp) 如下：

namespace webrtc { namespace internal {AudioState::AudioState(const AudioState::Config& config): config_(config),audio_transport_(config_.audio_mixer, config_.audio_processing.get()) {process_thread_checker_.Detach();RTC_DCHECK(config_.audio_mixer);RTC_DCHECK(config_.audio_device_module); }AudioState::~AudioState() {RTC_DCHECK(thread_checker_.IsCurrent());RTC_DCHECK(receiving_streams_.empty());RTC_DCHECK(sending_streams_.empty()); }AudioProcessing* AudioState::audio_processing() {RTC_DCHECK(config_.audio_processing);return config_.audio_processing.get(); }AudioTransport* AudioState::audio_transport() {return &audio_transport_; }bool AudioState::typing_noise_detected() const {RTC_DCHECK(thread_checker_.IsCurrent());return audio_transport_.typing_noise_detected(); }void AudioState::AddReceivingStream(webrtc::AudioReceiveStream* stream) {RTC_DCHECK(thread_checker_.IsCurrent());RTC_DCHECK_EQ(0, receiving_streams_.count(stream));receiving_streams_.insert(stream);if (!config_.audio_mixer->AddSource(static_cast<internal::AudioReceiveStream*>(stream))) {RTC_DLOG(LS_ERROR) << "Failed to add source to mixer.";}// Make sure playback is initialized; start playing if enabled.auto* adm = config_.audio_device_module.get();if (!adm->Playing()) {if (adm->InitPlayout() == 0) {if (playout_enabled_) {adm->StartPlayout();}} else {RTC_DLOG_F(LS_ERROR) << "Failed to initialize playout.";}} }void AudioState::RemoveReceivingStream(webrtc::AudioReceiveStream* stream) {RTC_DCHECK(thread_checker_.IsCurrent());auto count = receiving_streams_.erase(stream);RTC_DCHECK_EQ(1, count);config_.audio_mixer->RemoveSource(static_cast<internal::AudioReceiveStream*>(stream));if (receiving_streams_.empty()) {config_.audio_device_module->StopPlayout();} }void AudioState::AddSendingStream(webrtc::AudioSendStream* stream,int sample_rate_hz,size_t num_channels) {RTC_DCHECK(thread_checker_.IsCurrent());auto& properties = sending_streams_[stream];properties.sample_rate_hz = sample_rate_hz;properties.num_channels = num_channels;UpdateAudioTransportWithSendingStreams();// Make sure recording is initialized; start recording if enabled.auto* adm = config_.audio_device_module.get();if (!adm->Recording()) {if (adm->InitRecording() == 0) {if (recording_enabled_) {adm->StartRecording();}} else {RTC_DLOG_F(LS_ERROR) << "Failed to initialize recording.";}} }void AudioState::RemoveSendingStream(webrtc::AudioSendStream* stream) {RTC_DCHECK(thread_checker_.IsCurrent());auto count = sending_streams_.erase(stream);RTC_DCHECK_EQ(1, count);UpdateAudioTransportWithSendingStreams();if (sending_streams_.empty()) {config_.audio_device_module->StopRecording();} }void AudioState::SetPlayout(bool enabled) {RTC_LOG(INFO) << "SetPlayout(" << enabled << ")";RTC_DCHECK(thread_checker_.IsCurrent());if (playout_enabled_ != enabled) {playout_enabled_ = enabled;if (enabled) {null_audio_poller_.reset();if (!receiving_streams_.empty()) {config_.audio_device_module->StartPlayout();}} else {config_.audio_device_module->StopPlayout();null_audio_poller_ =absl::make_unique<NullAudioPoller>(&audio_transport_);}} }void AudioState::SetRecording(bool enabled) {RTC_LOG(INFO) << "SetRecording(" << enabled << ")";RTC_DCHECK(thread_checker_.IsCurrent());if (recording_enabled_ != enabled) {recording_enabled_ = enabled;if (enabled) {if (!sending_streams_.empty()) {config_.audio_device_module->StartRecording();}} else {config_.audio_device_module->StopRecording();}} }AudioState::Stats AudioState::GetAudioInputStats() const {RTC_DCHECK(thread_checker_.IsCurrent());const voe::AudioLevel& audio_level = audio_transport_.audio_level();Stats result;result.audio_level = audio_level.LevelFullRange();RTC_DCHECK_LE(0, result.audio_level);RTC_DCHECK_GE(32767, result.audio_level);result.total_energy = audio_level.TotalEnergy();result.total_duration = audio_level.TotalDuration();return result; }void AudioState::SetStereoChannelSwapping(bool enable) {RTC_DCHECK(thread_checker_.IsCurrent());audio_transport_.SetStereoChannelSwapping(enable); }void AudioState::UpdateAudioTransportWithSendingStreams() {RTC_DCHECK(thread_checker_.IsCurrent());std::vector<webrtc::AudioSendStream*> sending_streams;int max_sample_rate_hz = 8000;size_t max_num_channels = 1;for (const auto& kv : sending_streams_) {sending_streams.push_back(kv.first);max_sample_rate_hz = std::max(max_sample_rate_hz, kv.second.sample_rate_hz);max_num_channels = std::max(max_num_channels, kv.second.num_channels);}audio_transport_.UpdateSendingStreams(std::move(sending_streams),max_sample_rate_hz, max_num_channels); } } // namespace internal

總結一下：

• WebRtcVoiceEngine::Init() 建立了 WebRTC 音頻處理的靜態流水線。
• AudioTransport 將音頻處理靜態流水線的各個節點粘起來。
• webrtc::Call 使用 AudioState 將音頻處理流水線的靜態部分和動態部分連接起來。

超強干貨來襲 云風專訪：近40年碼齡，通宵達旦的技術人生

總結

以上是生活随笔為你收集整理的WebRTC 的音频处理流水线的全部內容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網站內容還不錯，歡迎將生活随笔推薦給好友。