windows平台，实现录音功能详解

Posted 2021-01-18 yuanchenhui

  音频处理分为播放和录音两类。对这些处理，微软提供了一些列函数，称之为Waveform Functions。这篇文章讨论录音功能。会对微软提供的函数做简单说明，并对这些函数封装成c++类，再进一步封装成c#类。

1 Waveform Functions函数简介

根据录音处理步骤，对这些函数做简单介绍。

  1.1  waveInOpen

MMRESULT waveInOpen(
   LPHWAVEIN       phwi,
   UINT            uDeviceID,
   LPCWAVEFORMATEX pwfx,
   DWORD_PTR       dwCallback,
   DWORD_PTR       dwCallbackInstance,
   DWORD           fdwOpen
);

pwfx为录音格式。普通对讲录音一般采样频率为8000HZ，位长为16bit，单声道。fdwOpen为回调类型，一般采用CALLBACK_FUNCTION，就是函数回调方式。当有录音设备打开、关闭、录音完成等事件发生时，系统会调用我们提供的回调函数。

1.2 waveInPrepareHeader，waveInAddBuffer
当录音设备打开后，需要你提供内存区域来存放录音数据。这两个函数就是完成这项功能。waveInPrepareHeader是准备内存，waveInAddBuffer是将内存加入到录音队列。
当录音完毕，会有回调通知，这时我们提供的内存中就存放着录音数据。回调函数是通过waveInOpen函数的dwCallbackInstance指定的。为了保持录音的连续性，录音队列要时时刻刻不能为空。录音队列的内存块个数一般要超过3个。就是第一次准备3个内存块。当有录音完毕，内存块个数会减1，这时我们立即补充一个内存块。

1.3 waveInStart，waveInStop 
这两个函数分别是启动和停止录音。一切准备完毕后，调用waveInStart，才会开始录音。

1.4 waveInClose
关闭录音。这个函数看起来非常简单，其实不然。这个函数会引发一些列事件，需要把这些事件处理好，否则会导致内存泄漏。当该函数被调用时，尚未存放录音的内存块会通过回调通知我们，这时需要将这些内存释放掉。

2 音频函数的c++封装
 封装目的就是隐藏细节，提供一种易于使用的调用模式。通过上文可以看到有几个细节难于处理：函数回调、内存块准备、内存释放。本类将这些细节隐藏，对外提供的模式为：
   打开录音设备--》启动录音--》不停轮询，读取已录音成功的数据--》关闭录音
 上述处理过程完全是线性化的。隐藏了数据准备、函数回调、内存释放等细节。封装类如下：
  头文件

#pragma once
#include "Mmsystem.h"
#include <list>
#include <queue>
#include "osType.h"

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

    BOOL IsOpen();

    void SetRecordDataLen(int len); //每个录音块长度
    BOOL Open(int nSamplesPerSec, int wBitsPerSample, int nChannels);
    void Close();

    BOOL WaitRecordedData(int waitMillisecond);
    int  GetRecordData(char* buffer, int bufferLen, int& bufferReadLen, int waitMillisecond = 0);

    BOOL StartRecord();
    BOOL StopRecord();
private:
    BOOL AddRecordBuffer();
    BOOL HaveRecordingBuffer();

    void AddToRecording(WAVEHDR *header);
    void RemoveFromRecording(WAVEHDR *header);

    void OnRcvRecordData(WAVEHDR * header);
    void AddToRecorded(WAVEHDR *header);
    void DelAllRecordData();

    void PrepareRecordData(int count);

    void OnClose();

private:
    BOOL            m_bInClosing;
    HWAVEIN            m_hWaveIn;
    WAVEFORMATEX    m_waveForm;
    int             m_recordBufferLen;

    std::list<WAVEHDR*> m_listWaveInRecording;
    CCritical m_recordingLock;

    std::queue<WAVEHDR*> m_listWaveInRecorded;
    CCritical    m_recordedLock;
    HANDLE        m_recordedDataEvent;

    static void WaveInProc(HWAVEOUT hwo, UINT uMsg, DWORD dwInstance, DWORD dwParam1, DWORD dwParam2);

};

实现文件

#include "stdafx.h"
#include "PcmRecord.h"

const int MaxDataCountInRecording = 10; //同时准备多少个 正在录音的buffer
void FreeWaveHeader(WAVEHDR *header);

PcmRecord::PcmRecord()
{
    m_hWaveIn = NULL;
    m_recordBufferLen = 1600;
    m_recordedDataEvent = CreateEvent(NULL, FALSE, FALSE, L"");
}


PcmRecord::~PcmRecord()
{
    Close();
}


void  PcmRecord::WaveInProc(HWAVEOUT hwo, UINT uMsg,
    DWORD dwInstance, DWORD dwParam1, DWORD dwParam2)
{
    PcmRecord *record = (PcmRecord*)dwInstance;
    if (uMsg == WOM_OPEN) //音频打开
    {
        return;
    }
    if (uMsg == WOM_CLOSE) //音频句柄关闭
    {
        record->OnClose();
        return;
    }

    if (uMsg == WIM_DATA)//获取了录制数据
    {
        WAVEHDR *header = (WAVEHDR*)dwParam1;
        record->OnRcvRecordData(header);
    }
}

void PcmRecord::OnClose()
{
    if (!m_bInClosing)
        Close();
}

void PcmRecord::OnRcvRecordData(WAVEHDR *header)
{
    //MMRESULT mmres = waveInUnprepareHeader(m_hWaveIn, header, sizeof(WAVEHDR));
    RemoveFromRecording(header);

    if (header->dwBytesRecorded > 0)
    {
        AddToRecorded(header);
    }
    else
    {
        FreeWaveHeader(header);
    }

    if (!m_bInClosing)
    {
        PrepareRecordData(MaxDataCountInRecording);
    }
}

void PcmRecord::AddToRecorded(WAVEHDR * header)
{
    {
        CCriticalLock  lock(m_recordedLock);
        m_listWaveInRecorded.push(header);
    }
    SetEvent(m_recordedDataEvent);
}

void PcmRecord::DelAllRecordData()
{
    CCriticalLock  lock(m_recordedLock);
    while (m_listWaveInRecorded.size() > 0)
    {
        WAVEHDR *header = m_listWaveInRecorded.front();
        m_listWaveInRecorded.pop();
        FreeWaveHeader(header);
    }
}

BOOL PcmRecord::WaitRecordedData(int waitMillisecond)
{
    {
        CCriticalLock  lock(m_recordedLock);
        if (m_listWaveInRecorded.size() > 0)
            return TRUE;
    }

    WaitForSingleObject(m_recordedDataEvent, waitMillisecond);

    {
        CCriticalLock  lock(m_recordedLock);
        return (m_listWaveInRecorded.size() > 0);
    }
}

int  PcmRecord::GetRecordData(char* buffer, int bufferLen,
    int& bufferReadLen, int waitMillisecond)
{
    bufferReadLen = 0;
    BOOL haveData;
    { // 因为有WaitForSingleObject调用，等待时间可能很长，所以要快速解锁
        CCriticalLock  lock(m_recordedLock);
        haveData = m_listWaveInRecorded.size() > 0;
    }

    if (!haveData
        && waitMillisecond == 0)
    {
        ResetEvent(m_recordedDataEvent);
        return 0;
    }

    //等待数据到来
    if (!haveData)
    {
        ResetEvent(m_recordedDataEvent);
        WaitForSingleObject(m_recordedDataEvent, waitMillisecond);
    }

    CCriticalLock  lock2(m_recordedLock);
    int copyIndex = 0;
    while ((bufferLen - copyIndex) >= m_recordBufferLen
        && m_listWaveInRecorded.size() > 0)
    {
        WAVEHDR *header = m_listWaveInRecorded.front();
        m_listWaveInRecorded.pop();
        memcpy(buffer, header->lpData, header->dwBytesRecorded);
        copyIndex += header->dwBytesRecorded;

        FreeWaveHeader(header);
    }

    bufferReadLen = copyIndex;
    return bufferReadLen;
}


BOOL PcmRecord::IsOpen()
{
    return m_hWaveIn != NULL;
}

BOOL PcmRecord::Open(int nSamplesPerSec, int wBitsPerSample, int nChannels)
{
    m_waveForm.nSamplesPerSec = nSamplesPerSec; /* sample rate */
    m_waveForm.wBitsPerSample = wBitsPerSample; /* sample size */
    m_waveForm.nChannels = nChannels; /* channels*/
    m_waveForm.cbSize = 0; /* size of _extra_ info */
    m_waveForm.wFormatTag = WAVE_FORMAT_PCM;
    m_waveForm.nBlockAlign = (m_waveForm.wBitsPerSample * m_waveForm.nChannels) >> 3;
    m_waveForm.nAvgBytesPerSec = m_waveForm.nBlockAlign * m_waveForm.nSamplesPerSec;

    MMRESULT mmres = waveInOpen(&m_hWaveIn, WAVE_MAPPER, &m_waveForm,
        (DWORD_PTR)WaveInProc, (DWORD_PTR)this, CALLBACK_FUNCTION);

    if (mmres != MMSYSERR_NOERROR)
    {
        return FALSE;
    }
    m_bInClosing = FALSE;
    PrepareRecordData(MaxDataCountInRecording);
    return TRUE;
}

void PcmRecord::SetRecordDataLen(int len)
{
    m_recordBufferLen = len;
}

 BOOL PcmRecord::StartRecord()
{
    MMRESULT mmres = waveInStart(m_hWaveIn);
    return (mmres == MMSYSERR_NOERROR);
}

 BOOL PcmRecord::StopRecord()
{
    MMRESULT mmres = waveInStop(m_hWaveIn);
    return (mmres == MMSYSERR_NOERROR);
}

void PcmRecord::Close()
{
    m_bInClosing = TRUE;
    MMRESULT mmres = waveInReset(m_hWaveIn);

    int n = 0;
    while (HaveRecordingBuffer() && n < 500)
    {
        Sleep(1);
        n++;
    }

    mmres = waveInClose(m_hWaveIn);
    m_hWaveIn = NULL;

    DelAllRecordData();
}

BOOL PcmRecord::HaveRecordingBuffer()
{
    CCriticalLock  lock(m_recordingLock);
    return m_listWaveInRecording.size() > 0;
}

void PcmRecord::AddToRecording(WAVEHDR *header)
{
    CCriticalLock  lock(m_recordingLock);
    m_listWaveInRecording.push_back(header);
}

void PcmRecord::RemoveFromRecording(WAVEHDR *header)
{
    CCriticalLock  lock(m_recordingLock);
    m_listWaveInRecording.remove(header);
}

void PcmRecord::PrepareRecordData(int count)
{
    CCriticalLock  lock(m_recordingLock);
    while (m_listWaveInRecording.size() < count)
    {
        if (!AddRecordBuffer())
            return;
    }
}

BOOL PcmRecord::AddRecordBuffer()
{
    WAVEHDR *header = new WAVEHDR();
    ZeroMemory(header, sizeof(WAVEHDR));
    //对应回调函数 DWORD_PTR dwParam1,
    header->dwUser = (DWORD_PTR)header;

    header->dwBufferLength = m_recordBufferLen;
    header->lpData = new char[m_recordBufferLen];

    MMRESULT result = waveInPrepareHeader(m_hWaveIn, header, sizeof(WAVEHDR));
    if (result != MMSYSERR_NOERROR)
    {
        FreeWaveHeader(header);
        return FALSE;
    }
    AddToRecording(header);

    result = waveInAddBuffer(m_hWaveIn, header, sizeof(WAVEHDR));
    if (result != MMSYSERR_NOERROR)
    {
        RemoveFromRecording(header);
        FreeWaveHeader(header);
        return FALSE;
    }

    return TRUE;
}

对于读取录音函数的使用特别说明一下。该函数定义如下：

int  PcmRecord::GetRecordData(char* buffer, int bufferLen,int& bufferReadLen, int waitMillisecond)

  bufferLen的长度要大于一个内存块。waitMillisecond为等待的毫秒数；当没有录音数据时，该函数会最多等待waitMillisecond毫秒。当waitMillisecond为0时，就是非阻塞调用。对于阻塞调用可以，采用独立线程读取；对于非阻塞，可以采用定时器方式轮询。

3 音频函数的c#封装

 在对c++类实现的基础上的进一步封装为c函数，可以供c#调用。这里的关键是c++函数封装为c函数。

    LIBPCMPLAY_API int64_t     PcmRecord_CreateHandle();
    LIBPCMPLAY_API void        PcmRecord_SetRecordDataLen(int64_t handle, int len);
    LIBPCMPLAY_API BOOL        PcmRecord_Open(int64_t handle, int nSamplesPerSec, int wBitsPerSample, int nChannels);
    LIBPCMPLAY_API BOOL        PcmRecord_IsOpen(int64_t handle);
    LIBPCMPLAY_API BOOL        PcmRecord_Start(int64_t handle);
    LIBPCMPLAY_API BOOL        PcmRecord_Stop(int64_t handle);
    LIBPCMPLAY_API int        PcmRecord_GetRecordData(int64_t handle, char* buffer, int bufferLen,
                                int& bufferReadLen, int waitMillisecond = 0);
    LIBPCMPLAY_API void        PcmRecord_Close(int64_t handle);

PcmRecord_CreateHandle 就是生成一个PcmRecord的实例，将该实例的指针返回。将handle定义为64位，这样32、64平台下处理方式就完全一样。handle就是类指针，这样就将复杂的类函数隐藏了。
实现函数

int64_t PcmPlay_CreateHandle()
{
    CPcmPlay *play = new CPcmPlay();
    return (int64_t)play;
}

BOOL PcmPlay_IsOpen(int64_t handle)
{
    CPcmPlay *play = (CPcmPlay*)handle;
    return play->IsOpen();
}

BOOL PcmPlay_Open(int64_t handle, int nSamplesPerSec,
    int wBitsPerSample, int nChannels)
{
    CPcmPlay *play = (CPcmPlay*)handle;
    return play->Open(nSamplesPerSec, wBitsPerSample, nChannels);
}

BOOL PcmPlay_SetVolume(int64_t handle, int volume)
{
    CPcmPlay *play = (CPcmPlay*)handle;
    return play->SetVolume(volume);
}

int  PcmPlay_Play(int64_t handle, char* block, int size)
{
    CPcmPlay *play = (CPcmPlay*)handle;
    return play->Play(block, size);
}

void PcmPlay_StopPlay(int64_t handle)
{
    CPcmPlay *play = (CPcmPlay*)handle;
    play->StopPlay();
}

BOOL PcmPlay_IsOnPlay(int64_t handle)
{
    CPcmPlay *play = (CPcmPlay*)handle;
    return play->IsOnPlay();
}

int64_t PcmPlay_GetLeftPlaySpan(int64_t handle)
{
    CPcmPlay *play = (CPcmPlay*)handle;
    return play->GetLeftPlaySpan();
}

int64_t PcmPlay_GetCurPlaySpan(int64_t handle)
{
    CPcmPlay *play = (CPcmPlay*)handle;
    return play->GetCurPlaySpan();
}

void PcmPlay_Close(int64_t handle)
{
    CPcmPlay *play = (CPcmPlay*)handle;
    play->Close();
}

void PcmPlay_CloseHandle(int64_t handle)
{
    CPcmPlay *play = (CPcmPlay*)handle;
    play->Close();
    delete play;
}

//录音 
int64_t    PcmRecord_CreateHandle()
{
    PcmRecord *record = new PcmRecord();
    return (int64_t)record;
}

void PcmRecord_SetRecordDataLen(int64_t handle, int len)
{
    PcmRecord *record = (PcmRecord*)handle;
    record->SetRecordDataLen(len);
}

BOOL PcmRecord_Open(int64_t handle, int nSamplesPerSec, int wBitsPerSample, int nChannels)
{
    PcmRecord *record = (PcmRecord*)handle;
    return record->Open(nSamplesPerSec,wBitsPerSample,nChannels);
}

BOOL PcmRecord_IsOpen(int64_t handle)
{
    PcmRecord *record = (PcmRecord*)handle;
    return record->IsOpen();
}

BOOL PcmRecord_Start(int64_t handle)
{
    PcmRecord *record = (PcmRecord*)handle;
    return record->StartRecord();
}

BOOL PcmRecord_Stop(int64_t handle)
{
    PcmRecord *record = (PcmRecord*)handle;
    return record->StopRecord();
}

int PcmRecord_GetRecordData(int64_t handle, char* buffer, int bufferLen,
    int& bufferReadLen, int waitMillisecond)
{
    PcmRecord *record = (PcmRecord*)handle;
    return record->GetRecordData(buffer, bufferLen, bufferReadLen, waitMillisecond);
}

void PcmRecord_Close(int64_t handle)
{
    PcmRecord *record = (PcmRecord*)handle;
    record->Close();
}

实现了对c语言的封装，下一步就是在c语言的基础上，封装成c#类。

　　class PcmRecord
    {
        long _handle = 0;
        int recordTimespan = 320;//每次录音长度 毫秒
        int bufferLenPerSample;

        public PcmRecord()
        {
            bufferLenPerSample = 16* recordTimespan;
        }

        ~PcmRecord()
        {
            if(_handle != 0)
            {
                PcmRecordWrapper.PcmRecord_Close(_handle);
            }
        }

        bool _isOpen = false;
        public bool IsOpen => _isOpen;
        public bool Open()
        {
            if (_isOpen)
                throw new Exception("先关闭，再打开！");

            _handle = PcmRecordWrapper.PcmRecord_CreateHandle();
            PcmRecordWrapper.PcmRecord_SetRecordDataLen(_handle, bufferLenPerSample);
            _isOpen = PcmRecordWrapper.PcmRecord_Open(_handle, 8000, 16, 1)==1;
            return true;
        }

        public bool Start()
        {
            if (!_isOpen)
                throw new Exception("录音设备还没打开！");
            return PcmRecordWrapper.PcmRecord_Start(_handle) == 1;
        }

        public bool Stop()
        {
            if (!_isOpen)
                throw new Exception("录音设备还没打开！");
            return PcmRecordWrapper.PcmRecord_Stop(_handle) == 1;
        }

        public byte[] GetPcmData(int waitMillisecond)
        {
            if (!_isOpen)
                throw new Exception("录音设备还没打开！");

            byte[] bufferRecord = new byte[bufferLenPerSample];
            GCHandle hinBuffer = GCHandle.Alloc(bufferRecord, GCHandleType.Pinned);

            byte[] readLen = new byte[4];
            GCHandle hinReadLen = GCHandle.Alloc(readLen, GCHandleType.Pinned);

            PcmRecordWrapper.PcmRecord_GetRecordData(_handle, hinBuffer.AddrOfPinnedObject(), bufferRecord.Length,
               hinReadLen.AddrOfPinnedObject(), waitMillisecond);
            hinBuffer.Free();
            hinReadLen.Free();

            int returnLen = BitConverter.ToInt32(readLen, 0);

            if (returnLen == 0)
                return null;
            if (returnLen == bufferRecord.Length)
                return bufferRecord;

            Array.Resize(ref bufferRecord, returnLen);
            return bufferRecord;
        }

        public void Close()
        {
            if (!_isOpen)
                return;
            PcmRecordWrapper.PcmRecord_Close(_handle);
            _handle = 0;
            _isOpen = false;
        }
    }

    public class PcmRecordWrapper
    {
        private const string DLLName = "LibPcmPlay.dll";

        [DllImport(DLLName, EntryPoint = "PcmRecord_CreateHandle", CallingConvention = CallingConvention.Cdecl)]
        public static extern long PcmRecord_CreateHandle();

        [DllImport(DLLName, EntryPoint = "PcmRecord_SetRecordDataLen", CallingConvention = CallingConvention.Cdecl)]
        public static extern void PcmRecord_SetRecordDataLen(long handle, int len);

        [DllImport(DLLName, EntryPoint = "PcmRecord_Open", CallingConvention = CallingConvention.Cdecl)]
        public static extern int PcmRecord_Open(long handle, int nSamplesPerSec, int wBitsPerSample, int nChannels);

        [DllImport(DLLName, EntryPoint = "PcmRecord_IsOpen", CallingConvention = CallingConvention.Cdecl)]
        public static extern int PcmRecord_IsOpen(long handle);

        [DllImport(DLLName, EntryPoint = "PcmRecord_Start", CallingConvention = CallingConvention.Cdecl)]
        public static extern int PcmRecord_Start(long handl);

        [DllImport(DLLName, EntryPoint = "PcmRecord_Stop", CallingConvention = CallingConvention.Cdecl)]
        public static extern int PcmRecord_Stop(long handl);

        [DllImport(DLLName, EntryPoint = "PcmRecord_GetRecordData", CallingConvention = CallingConvention.Cdecl)]
        public static extern int PcmRecord_GetRecordData(long handle, IntPtr buffer, Int32 bufferLen,
                            IntPtr bufferReadLen, int waitMillisecond);

        [DllImport(DLLName, EntryPoint = "PcmRecord_Close", CallingConvention = CallingConvention.Cdecl)]
        public static extern void PcmRecord_Close(long handl);

    }
}

总结：windows平台为我们提供了录音相关函数。平台提供的函数考虑到各种应用场景，使用起来非常灵活，但是容易出错，需要关注细节。合适的才是最好的。根据自身的需求，选用一种合适的处理模式；这种模式既能满足我们的功能需求，又要易于使用。本文不仅提供了对录音函数的封装，也提供一种处理复杂问题的思路。希望能抛砖引玉！