[转]directsound抓取麦克风PCM数据封装类

网上有很多方法从麦克风读取PCM数据，不想一一举例。只是在这里发布一个我自己写的directsound的麦克风PCM数据采集类，通过它，可以很方便的利用directsound技术把麦克风的数据采集到，而且本身，开发者不必太在意自己会不会directsound编程，可以很方便的让开发者的主要精力集中于程序本身，而不是细节。就是需要安装directx9b的sdk，然后在编译的时候加入以下lib： strmiids.lib Quartz.lib winmm.lib dsound.lib dxguid.lib

这个是头文件：

#pragma once
#ifndef _CAPTURE_SOUND_H_
#define _CAPTURE_SOUND_H_
#include <mmsystem.h>
#include <dsound.h>
#define NUM_REC_NOTIFICATIONS  16
class CAdoFrameHandler {
public:
 virtual void AdoFrameData(BYTE* pBuffer, long lBufferSize) =  ;
};
class CCaptureAudio
{
public:
 BOOL        m_bRecording ;  //recording now ? also used by event recv thread
protected:
 LPDIRECTSOUNDCAPTURE8    m_pCapDev ;   //capture device ptr
 LPDIRECTSOUNDCAPTUREBUFFER m_pCapBuf ;   //capture loop buffer ptr
 LPDIRECTSOUNDNOTIFY8    m_pNotify ;   //capture auto-notify event callback handler ptr
 GUID        m_guidCapDevId ;  //capture device id
 WAVEFORMATEX      m_wfxInput;   //input wave format description struct
 DSBPOSITIONNOTIFY     m_aPosNotify[NUM_REC_NOTIFICATIONS + ]; //notify flag array
 HANDLE        m_hNotifyEvent;   //notify event
 BOOL        m_abInputFmtSupported[];
 DWORD        m_dwCapBufSize;  //capture loop buffer size
 DWORD        m_dwNextCapOffset;//offset in loop buffer
 DWORD        m_dwNotifySize;  //notify pos when loop buffer need to emit the event
 CAdoFrameHandler*     m_frame_handler ; // outer frame data dealer ptr
public: // callback func to add enum devices string name
 static BOOL CALLBACK enum_dev_proc(LPGUID lpGUID, LPCTSTR lpszDesc,
            LPCTSTR lpszDrvName, LPVOID lpContext ) ;
 static UINT notify_capture_thd(LPVOID data) ;
protected:
 HRESULT InitDirectSound(GUID dev_id = GUID_NULL) ;
 HRESULT FreeDirectSound() ;
 HRESULT InitNotifications() ;
 HRESULT CreateCaptureBuffer(WAVEFORMATEX * wfx) ;
 HRESULT StartOrStopRecord(BOOL bStartRec) ;
 HRESULT RecordCapturedData() ;
 void    SetWavFormat(WAVEFORMATEX * wfx) ;
public:
 CCaptureAudio(void);
 ~CCaptureAudio(void);
 BOOL EnumDevices(HWND hList) ;
 BOOL Open(void) ;
 BOOL Close() ;
 void GrabAudioFrames(BOOL bGrabAudioFrames, CAdoFrameHandler* frame_handler) ;
};
#endif

下面这个是cpp文件：

#include "StdAfx.h"
#include ".\captureaudio.h"
#include <mmsystem.h>
#include <dsound.h>
#ifndef SAFE_DELETE
#define SAFE_DELETE(p)  { if(p) { delete (p);     (p)=NULL; } }
#endif
#ifndef SAFE_RELEASE
#define SAFE_RELEASE(p) { if(p) { (p)->Release(); (p)=NULL; } }
#endif
#ifndef MAX
#define MAX(a,b)        ( (a) > (b) ? (a) : (b) )
#endif
CCaptureAudio::CCaptureAudio(void)
{
 if(FAILED(CoInitialize(NULL))) /*, COINIT_APARTMENTTHREADED)))*/
 {
  AfxMessageBox("CCaptureAudio CoInitialize Failed!\r\n");
  return;
 }
 m_pCapDev = NULL ;
 m_pCapBuf = NULL ;
 m_pNotify = NULL ;
 // set default wave format PCM
 ZeroMemory( &m_wfxInput, sizeof(m_wfxInput));
 m_wfxInput.wFormatTag = WAVE_FORMAT_PCM;
 m_guidCapDevId = GUID_NULL ;
 m_bRecording = FALSE ;
 m_hNotifyEvent = NULL ;
}
CCaptureAudio::~CCaptureAudio(void)
{
 CoUninitialize() ;
}
BOOL CALLBACK CCaptureAudio::enum_dev_proc(LPGUID lpGUID, LPCTSTR lpszDesc,
             LPCTSTR lpszDrvName, LPVOID lpContext)
{
 HWND hList = (HWND)lpContext;
 if(!hList) return FALSE ;
 LPGUID lpTemp = NULL;
 if (lpGUID != NULL) {
  // NULL only for "Primary Sound Driver".
  if ((lpTemp = (LPGUID)malloc(sizeof(GUID))) == NULL) return(TRUE);
  memcpy(lpTemp, lpGUID, sizeof(GUID));
 }
 ::SendMessage(hList, CB_ADDSTRING, ,(LPARAM)lpszDesc);
 ::SendMessage(hList, LB_SETITEMDATA, , (LPARAM)lpTemp) ;
 free(lpTemp);
 return(TRUE);
}
UINT CCaptureAudio::notify_capture_thd(LPVOID data)
{
 CCaptureAudio * pado = static_cast<CCaptureAudio *>(data) ;
 MSG   msg;
 HRESULT hr ;
 DWORD dwResult ;
 while(pado->m_bRecording) {
  dwResult = MsgWaitForMultipleObjects( , &(pado->m_hNotifyEvent), FALSE, INFINITE, QS_ALLEVENTS );
  switch( dwResult ) {
  case WAIT_OBJECT_0 + :
   // g_hNotificationEvents[0] is signaled
   // This means that DirectSound just finished playing
   // a piece of the buffer, so we need to fill the circular
   // buffer with new sound from the wav file
   if( FAILED( hr = pado->RecordCapturedData() ) ) {
    AfxMessageBox("Error handling DirectSound notifications.") ;
    pado->m_bRecording = FALSE ;
   }
   break;
  case WAIT_OBJECT_0 + :
   // Windows messages are available
   while( PeekMessage( &msg, NULL, , , PM_REMOVE ) ) {
     TranslateMessage( &msg );
     DispatchMessage( &msg );
    if( msg.message == WM_QUIT ) pado->m_bRecording = FALSE ;
   }
   break;
  }
 }
 AfxEndThread(, TRUE) ;
 return  ;
}
BOOL CCaptureAudio::EnumDevices(HWND hList)
{
 if (FAILED(DirectSoundCaptureEnumerate (
   (LPDSENUMCALLBACK)(CCaptureAudio::enum_dev_proc),
   (VOID*)&hList)))
 {
  return(FALSE);
 }
 return (TRUE) ;
}
BOOL CCaptureAudio::Open(void)
{
 HRESULT hr ;
 if(!m_bRecording) {
  hr = InitDirectSound() ;
 }
 return (FAILED(hr)) ? FALSE : TRUE ;
}
BOOL CCaptureAudio::Close()
{
 HRESULT hr ;
 hr = FreeDirectSound() ;
 CloseHandle(m_hNotifyEvent) ;
 return (FAILED(hr)) ? FALSE : TRUE ;
}
HRESULT CCaptureAudio::InitDirectSound(GUID dev_id)
{
 HRESULT hr ;
 m_guidCapDevId = dev_id ;
 ZeroMemory( &m_aPosNotify, sizeof(DSBPOSITIONNOTIFY) * (NUM_REC_NOTIFICATIONS + ) ) ;
 m_dwCapBufSize =  ;
 m_dwNotifySize =  ;
 // Create IDirectSoundCapture using the preferred capture device
 hr = DirectSoundCaptureCreate(&m_guidCapDevId, &m_pCapDev, NULL ) ;
 // init wave format
 SetWavFormat(&m_wfxInput) ;
 return (FAILED(hr)) ? S_FALSE : S_OK ;
}
HRESULT CCaptureAudio::FreeDirectSound()
{
 // Release DirectSound interfaces
 SAFE_RELEASE( m_pNotify ) ;
 SAFE_RELEASE( m_pCapBuf ) ;
 SAFE_RELEASE( m_pCapDev ) ;
 return S_OK;
}
HRESULT CCaptureAudio::CreateCaptureBuffer(WAVEFORMATEX * wfx)
{
 HRESULT hr;
 DSCBUFFERDESC dscbd;
 SAFE_RELEASE( m_pNotify );
 SAFE_RELEASE( m_pCapBuf );
 // Set the notification size
 m_dwNotifySize = MAX( , wfx->nAvgBytesPerSec /  ) ;
 m_dwNotifySize -= m_dwNotifySize % wfx->nBlockAlign ;
 // Set the buffer sizes
 m_dwCapBufSize = m_dwNotifySize * NUM_REC_NOTIFICATIONS;
 SAFE_RELEASE( m_pNotify );
 SAFE_RELEASE( m_pCapBuf );
 // Create the capture buffer
 ZeroMemory( &dscbd, sizeof(dscbd) );
 dscbd.dwSize        = sizeof(dscbd);
 dscbd.dwBufferBytes = m_dwCapBufSize;
 dscbd.lpwfxFormat   = wfx ; // Set the format during creatation
 if( FAILED( hr = m_pCapDev->CreateCaptureBuffer( &dscbd, &m_pCapBuf, NULL ) ) )
  return S_FALSE ;
 m_dwNextCapOffset = ;
 if( FAILED( hr = InitNotifications() ) )
  return S_FALSE ;
 return S_OK;
}
HRESULT CCaptureAudio::InitNotifications()
{
 HRESULT hr;
 int i ;
 if( NULL == m_pCapBuf )
  return S_FALSE;
 // create auto notify event
 m_hNotifyEvent = CreateEvent( NULL, FALSE, FALSE, NULL );
 // Create a notification event, for when the sound stops playing
 if( FAILED( hr = m_pCapBuf->QueryInterface( IID_IDirectSoundNotify, (VOID**)&m_pNotify ) ) )
  return S_FALSE ;
 // Setup the notification positions
 for( i = ; i < NUM_REC_NOTIFICATIONS; i++ ) {
  m_aPosNotify[i].dwOffset = (m_dwNotifySize * i) + m_dwNotifySize - ;
  m_aPosNotify[i].hEventNotify = m_hNotifyEvent;
 }
 // Tell DirectSound when to notify us. the notification will come in the from
 // of signaled events that are handled in WinMain()
 if( FAILED( hr = m_pNotify->SetNotificationPositions( NUM_REC_NOTIFICATIONS, m_aPosNotify ) ) )
  return S_FALSE ;
 return S_OK;
}
HRESULT CCaptureAudio::StartOrStopRecord(BOOL bStartRec)
{
 HRESULT hr;
 if( bStartRec ) {
  // Create a capture buffer, and tell the capture
  // buffer to start recording
  if( FAILED( hr = CreateCaptureBuffer( &m_wfxInput ) ) )
   return S_FALSE ;
  if( FAILED( hr = m_pCapBuf->Start( DSCBSTART_LOOPING ) ) )
   return S_FALSE ;
  // create notify event recv thread
  AfxBeginThread(CCaptureAudio::notify_capture_thd, (LPVOID)(this)) ;
 } else {
  // Stop the capture and read any data that
  // was not caught by a notification
  if( NULL == m_pCapBuf )
   return S_OK;
  // wait until the notify_event_thd thread exit and release the resources.
  Sleep() ;
  // Stop the buffer, and read any data that was not
  // caught by a notification
  if( FAILED( hr = m_pCapBuf->Stop() ) )
   return S_OK ;
  if( FAILED( hr = RecordCapturedData() ) )
   return S_FALSE ;
 }
 return S_OK;
}
HRESULT CCaptureAudio::RecordCapturedData()
{
 HRESULT hr;
 VOID*   pbCaptureData    = NULL;
 DWORD   dwCaptureLength;
 VOID*   pbCaptureData2   = NULL;
 DWORD   dwCaptureLength2;
 DWORD   dwReadPos;
 DWORD   dwCapturePos;
 LONG lLockSize;
 if( NULL == m_pCapBuf )
  return S_FALSE; 

 if( FAILED( hr = m_pCapBuf->GetCurrentPosition( &dwCapturePos, &dwReadPos ) ) )
  return S_FALSE;
 lLockSize = dwReadPos - m_dwNextCapOffset;
 if( lLockSize <  )
  lLockSize += m_dwCapBufSize;
 // Block align lock size so that we are always write on a boundary
 lLockSize -= (lLockSize % m_dwNotifySize);
 if( lLockSize ==  )
  return S_FALSE;
 // Lock the capture buffer down
 if( FAILED( hr = m_pCapBuf->Lock( m_dwNextCapOffset, lLockSize,
           &pbCaptureData, &dwCaptureLength,
           &pbCaptureData2, &dwCaptureLength2, 0L ) ) )
  return S_FALSE ;
 // call the outer data handler
 if(m_frame_handler) {
  m_frame_handler->AdoFrameData((BYTE*)pbCaptureData, dwCaptureLength) ;
 }

 // Move the capture offset along
 m_dwNextCapOffset += dwCaptureLength;
 m_dwNextCapOffset %= m_dwCapBufSize; // Circular buffer
 if( pbCaptureData2 != NULL ) {
  // call the outer data handler
  if(m_frame_handler) {
   m_frame_handler->AdoFrameData((BYTE*)pbCaptureData, dwCaptureLength) ;
  }
  // Move the capture offset along
  m_dwNextCapOffset += dwCaptureLength2;
  m_dwNextCapOffset %= m_dwCapBufSize; // Circular buffer
 }
 // Unlock the capture buffer
 m_pCapBuf->Unlock( pbCaptureData,  dwCaptureLength, pbCaptureData2, dwCaptureLength2 );
 return S_OK;
}
void CCaptureAudio::SetWavFormat(WAVEFORMATEX * wfx)
{
 // get the default capture wave formate
 ZeroMemory(wfx, sizeof(WAVEFORMATEX)) ;
 wfx->wFormatTag = WAVE_FORMAT_PCM;
 // 8KHz, 16 bits PCM, Mono
 wfx->nSamplesPerSec =  ;
 wfx->wBitsPerSample =  ;
 wfx->nChannels  =  ;
 wfx->nBlockAlign = wfx->nChannels * ( wfx->wBitsPerSample /  ) ;
 wfx->nAvgBytesPerSec = wfx->nBlockAlign * wfx->nSamplesPerSec;
}
void CCaptureAudio::GrabAudioFrames(BOOL bGrabAudioFrames, CAdoFrameHandler* frame_handler)
{
 m_frame_handler = frame_handler ;
 m_bRecording = bGrabAudioFrames ;
 StartOrStopRecord(m_bRecording) ;
}

使用的时候，也很简单，我这里声明了一个纯虚类CAdoFrameHandler，这个类专门是用来让使用者重载它的纯虚函数的，只要重载了以后，设置正确，就可以自动开始采集音频数据了。注意，在这个类里面，我用的是8KHz,16Bits,Mono单声道的PCM数据采集。

使用的时候，首先：

#include "CaptureAudio.h"

然后：

class CMyClass : public CAdoFrameHandler {

public: // override the CAdoFrameHandler
 void AdoFrameData(BYTE* pBuffer, long lBufferSize) ;  // 这个类重载一下，就可以采集了

protected:
  CCaptureAudio   m_cap_ado ; // 这个对象就是用来采集音频数据的
} ;

在OnInitDialog类中，我们可以使用如下初始化方法：

 

打开mic，同时初始化directsound：

m_cap_ado.Open() ;

开始采集声音就是：

m_cap_ado.GrabAudioFrames(TRUE, this) ;

调用它以后，只要你重载了上面的那个函数，directsound就会周期性的从麦克采集数据，然后调用该函数。

 

停止声音采集是：

m_cap_ado.GrabAudioFrames(FALSE, NULL) ;

关闭mic，同时释放directsound：

m_cap_ado.Close() ;

就这么简单的几步，就可以完成麦克风的音频数据采集。

转自：http://blog.chinaunix.net/uid-8272118-id-2033248.html