使用 Portaudio C++ 进行实时音频处理

Question

我正在尝试制作一个 C++ 应用程序，通过 VoIP 协议在 2 个客户端（使用 UDP）之间传输音频。

我正在使用 Portaudio C 库，但在封装这个库时遇到了问题。 为了将录制的音频发送到另一个客户端，我想在录制时（实时）获取声音样本。

目前我只能录制声音，然后播放我录制的内容。 我对这个图书馆一点都不舒服，任何帮助将不胜感激。

这是我到目前为止所做的。

Audio.cpp -> 回调方法：

static int PaRecordCallback(const void *input, void *output, unsigned long frameCount, \
    const PaStreamCallbackTimeInfo *timeInfo, PaStreamCallbackFlags statusFlags, void *userData)
{
    Audio *audio = reinterpret_cast<Audio *>(userData);

    return audio->RecordCallback(input, output, frameCount, timeInfo, statusFlags);
}

static int PaPlayCallback(const void *input, void *output, unsigned long frameCount, \
    const PaStreamCallbackTimeInfo *timeInfo, PaStreamCallbackFlags statusFlags, void *userData)
{
    Audio *audio = reinterpret_cast<Audio *>(userData);

    return audio->PlayCallback(input, output, frameCount, timeInfo, statusFlags);
}

int Audio::RecordCallback(const void *input, void *output, unsigned long &frameCount, \
    const PaStreamCallbackTimeInfo *timeInfo, PaStreamCallbackFlags &statusFlags)
{
    std::cout << "Frame index:\t\t" << _recordedFrameIndex << std::endl << "Max frame index:\t" << _maxFrameIndex << std::endl << "--------------" << std::endl;
    const SAMPLE *rptr = static_cast<const SAMPLE *>(input);
    SAMPLE *wptr = &_recordedSamples[_recordedFrameIndex * NUM_CHANNELS];
    unsigned long framesLeft = _maxFrameIndex - _recordedFrameIndex;
    unsigned long framesToCalc;
    int finished;

    if (framesLeft < frameCount) {
        framesToCalc = framesLeft;
        finished = paComplete;
    } else {
        framesToCalc = frameCount;
        finished = paContinue;
    }
    for (unsigned long i = 0; i < framesToCalc; i++) {
        *wptr++ = *rptr++;
        if (NUM_CHANNELS == 2)
            *wptr++ = *rptr++;
    }
    _recordedFrameIndex += framesToCalc;
    return finished;
}

int Audio::PlayCallback(const void *input, void *output, unsigned long &frameCount, \
    const PaStreamCallbackTimeInfo *timeInfo, PaStreamCallbackFlags &statusFlags)
{
    SAMPLE *rptr = &_recordedSamples[_playedFrameIndex * NUM_CHANNELS];
    SAMPLE *wptr = static_cast<SAMPLE *>(output);
    unsigned long framesLeft = _maxFrameIndex - _playedFrameIndex;
    unsigned int i;
    int finished;

    if (framesLeft < frameCount) {
        for (i = 0; i < framesLeft; i++) {
            *wptr++ = *rptr++;
            if (NUM_CHANNELS == 2)
                *wptr++ = *rptr++;
        }
        for (; i < frameCount; i++) {
            *wptr++ = 0;
            if (NUM_CHANNELS == 2)
                *wptr++ = 0;
        }
        _playedFrameIndex += framesLeft;
        finished = paComplete;
    } else {
        for (i = 0; i < frameCount; i++) {
            *wptr++ = *rptr++;
            if (NUM_CHANNELS == 2)
                *wptr++ = *rptr++;
        }
        _playedFrameIndex += frameCount;
        finished = paContinue;
    }
    return finished;
}

Audio.cpp -> 录制和播放方法：

void Audio::Record()
{
    if (!_recordStream) {
        OpenRecordStream();
        _recordedFrameIndex = 0;
        _err = Pa_StartStream(_recordStream);
        if (_err != paNoError)
            AudioError("Audio::Record -> Pa_StartStream()");
        std::cout << "Audio record stream started." << std::endl;
        std::cout << "Recording ..." << std::endl;
        _recording = true;
        fflush(stdout);
    } else if (_recording)
        Pa_Sleep(1000);
}

void Audio::Play()
{
    if (!_playStream) {
        OpenPlayStream();
        _playedFrameIndex = 0;
        _err = Pa_StartStream(_playStream);
        if (_err != paNoError)
            AudioError("Audio::Play -> Pa_StartStream()");
        std::cout << "Audio play stream started." << std::endl;
        std::cout << "Playing ..." << std::endl;
        _playing = true;
        fflush(stdout);
    } else if (_playing)
        Pa_Sleep(500);
}

Audio.hpp -> 类音频：

#include <portaudio.h>

typedef int16_t SAMPLE;

#define PA_SAMPLE_TYPE      paInt16
#define PRINTF_S_FORMAT     "%.8f"
#define SAMPLE_RATE         44100
#define SAMPLE_SILENCE      0.0f
#define FRAMES_PER_BUFFER   1
#define NUM_SECONDS         5
#define NUM_CHANNELS        2
#define DITHER_FLAG         0
#define WRITE_TO_FILE       0
#define SAMPLE_SIZE         NUM_SECONDS * SAMPLE_RATE * NUM_CHANNELS

class Audio
{
    public:
        Audio();
        ~Audio();

        void Record();
        void Play();

        void OpenRecordStream();
        void OpenPlayStream();

        void CloseRecordStream();
        void ClosePlayStream();

        const bool &isRecording() const;
        const bool &isPlaying() const;

        const PaStream *GetRecordStream() const;
        const PaStream *GetPlayStream() const;

        void GetSamples(SAMPLE *);
        void SetSamples(SAMPLE *);

        int RecordCallback(const void *, void *, unsigned long &, \
            const PaStreamCallbackTimeInfo *, PaStreamCallbackFlags &);
        int PlayCallback(const void *, void *, unsigned long &, \
            const PaStreamCallbackTimeInfo *, PaStreamCallbackFlags &);
        bool _recording;
        bool _playing;

    protected:
    private:
        // Functions:
        void AudioError(const std::string &);

        // Variables:
        PaError _err;
        PaStream *_playStream;
        PaStream *_recordStream;
        SAMPLE *_samplesToPlay;
        SAMPLE *_recordedSamples;
        unsigned long _recordedFrameIndex;
        unsigned long _playedFrameIndex;
        unsigned long _maxFrameIndex;
        PaStreamParameters _inputParameters;
        PaStreamParameters _outputParameters;
};

如果它很长，我深表歉意，但我希望您拥有所有必要的信息来理解我的问题。 我不经常问问题，所以我真的需要一些帮助。

谢谢你。

Answer 1

您需要使用全双工回调来实时录制和播放，以便您以块的形式捕获录音的输入缓冲区（当您想要录制时）并将录音（甚至传入的声音）发送到输出缓冲区，也在大块。 块大小通常为 64 到 4096 帧，每帧通常包含 2 个样本（通道 L 和通道 R）

全双工回调是一种循环缓冲区，当输入缓冲区被 ADC 填充时，它会为您带来 x 帧，并在其中用 x 帧填充输出缓冲区，以便在 DAC 请求时准备就绪。