directsound抓取麥克風PCM資料封裝類

directsound抓取麥克風PCM資料封裝類

網上有很多方法從麥克風讀取PCM資料，不想一一舉例。只是在這裡發布一個我自己寫的directsound的麥克風PCM資料擷取類，通過它，可以很方便的利用directsound技術把麥克風的資料擷取到，而且本身，開發人員不必太在意自己會不會directsound編程，可以很方便的讓開發人員的主要精力集中於程式本身，而不是細節。這個是標頭檔：#pragma once #ifndef _CAPTURE_SOUND_H_ #define _CAPTURE_SOUND_H_ #include <mmsystem.h> #include <dsound.h>#define NUM_REC_NOTIFICATIONS  16class CAdoFrameHandler { public:  virtual void AdoFrameData(BYTE* pBuffer, long lBufferSize) = 0 ; }; 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 + 1]; //notify flag array  HANDLE        m_hNotifyEvent;   //notify event  BOOL        m_abInputFmtSupported[20];  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 下面這個是cpp檔案：#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, 0,(LPARAM)lpszDesc);  ::SendMessage(hList, LB_SETITEMDATA, 0, (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( 1, &(pado->m_hNotifyEvent), FALSE, INFINITE, QS_ALLEVENTS );   switch( dwResult ) {   case WAIT_OBJECT_0 + 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 + 1:    // Windows messages are available    while( PeekMessage( &msg, NULL, 0, 0, PM_REMOVE ) ) {      TranslateMessage( &msg );      DispatchMessage( &msg );     if( msg.message == WM_QUIT ) pado->m_bRecording = FALSE ;    }    break;   }  }  AfxEndThread(0, TRUE) ;  return 0 ; }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 + 1) ) ;  m_dwCapBufSize = 0 ;  m_dwNotifySize = 0 ; // 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( 1024, wfx->nAvgBytesPerSec / 8 ) ;  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 = 0; 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 = 0; i < NUM_REC_NOTIFICATIONS; i++ ) {   m_aPosNotify[i].dwOffset = (m_dwNotifySize * i) + m_dwNotifySize - 1;   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(500) ;  // 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 < 0 )   lLockSize += m_dwCapBufSize; // Block align lock size so that we are always write on a boundary  lLockSize -= (lLockSize % m_dwNotifySize); if( lLockSize == 0 )   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 = 8000 ;  wfx->wBitsPerSample = 16 ;  wfx->nChannels  = 1 ; wfx->nBlockAlign = wfx->nChannels * ( wfx->wBitsPerSample / 8 ) ;  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) ; } 使用的時候，也很簡單，我這裡聲明了一個純虛類CAdoFrameHandler，這個類專門是用來讓使用者重載它的純虛函數的，只要重載了以後，設定正確，就可以自動開始採集音頻資料了。注意，在這個類裡面，我用的是8KHz,16Bits,Mono單聲道的PCM資料擷取。使用的時候，首先：#include "CaptureAudio.h"然後：class CMyClass : public CAdoFrameHandler {...public: // override the CAdoFrameHandler  void AdoFrameData(BYTE* pBuffer, long lBufferSize) ;  // 這個類重載一下，就可以採集了 protected:  CCaptureAudio   m_cap_ado ; // 這個對象就是用來採集音頻資料的} ;  在OnInitDialog類中，我們可以使用如下初始化方法： 開啟mic，同時初始化directsound：m_cap_ado.Open() ;  開始採集聲音就是：m_cap_ado.GrabAudioFrames(TRUE, this) ; 調用它以後，只要你重載了上面的那個函數，directsound就會周期性的從麥克採集資料，然後調用該函數。 停止聲音採集是：m_cap_ado.GrabAudioFrames(FALSE, NULL) ; 關閉mic，同時釋放directsound：m_cap_ado.Close() ;  就這麼簡單的幾步，就可以完成麥克風的音頻資料擷取。有問題歡迎大家多多交流。