语音识别与朗读 with MS SAPI5.1

http://blog.csdn.net/In355Hz/archive/2003/07/04/5142.aspx

Hello World by Microsoft Speech SDK 5.1

     刚刚从 Microsoft 网站下栽了 Speech SDK 5.1 和中日文发声补丁，参考 Speech SDK文档写了几个小程序，尽管我也是刚刚接触 Speech若干小时的初学者，还是拿出来共享吧：

 

Speech SDK 采用 COM 的形式封装，和大多数COM对象一样，下面是一个似曾相识的初始化：

 

    CoInitialize(NULL);

 

    CLSIDFromProgID(L"SAPI.SpVoice", &CLSID_SpVoice);

 

    ISpVoice * pSpVoice = NULL;

 

    if (FAILED(CoCreateInstance(CLSID_SpVoice, NULL,

        CLSCTX_INPROC_SERVER, IID_ISpVoice, (void **)&pSpVoice)))

    {

        cout << "Failed to create instance of ISpVoice!" << endl;

        return -1;

    }

 

    现在我们用它来说 “Hello World”，非常简单：

 

pSpVoice->Speak(L"Hello World!", SPF_DEFAULT, NULL);

 

最后，清除这一切

 

pSpVoice->Release();

 

CoUninitialize();

 

很容易，是不是？下面我们稍微改变一下：

 

IEnumSpObjectTokens *pSpEnumTokens = NULL;

if (SUCCEEDED(SpEnumTokens(SPCAT_VOICES, L"language=409", NULL, &pSpEnumTokens)))

{

       ISpObjectToken *pSpToken = NULL;

 

       while (SUCCEEDED(pSpEnumTokens->Next(1, &pSpToken, NULL)) &&

              pSpToken != NULL)

       {

              pSpVoice->SetVoice(pSpToken);

              pSpVoice->Speak(L"Hello World!", SPF_DEFAULT, NULL);

 

              pSpToken->Release();

       }

 

       pSpEnumTokens->Release();

}

 

如果只需要一种 Voice，SpFindBestToken 更加方便：

 

ISpObjectToken * pSpObjectToken = NULL;

 

if (SUCCEEDED(SpFindBestToken(SPCAT_VOICES, L"gender=female",

       NULL, &pSpObjectToken)))

{

       pSpVoice->SetVoice(pSpObjectToken);

 

       pSpVoice->Speak(L"Hello World!", SPF_DEFAULT, NULL);

 

       pSpObjectToken->Release();

}

 

现在改为使用中文：

 

ISpObjectToken * pSpObjectToken = NULL;

 

if (SUCCEEDED(SpFindBestToken(SPCAT_VOICES, L"language=804",

       NULL, &pSpObjectToken)))

{

       pSpVoice->SetVoice(pSpObjectToken);

 

       pSpVoice->Speak(L"世界你好！", SPF_DEFAULT, NULL);

 

       pSpObjectToken->Release();

}

 

ISpVoice::Speak 可以允许朗读的句子中包含 XML 标记，其实以上这一切用 XML 标记将变得十分简单：

 

pSpVoice->Speak(L"<lang langid=/"804/">世界你好！</lang>", SPF_DEFAULT, NULL);

 

    pSpVoice->Speak(L"<voice required=/"gender=female/">Hello World!</voice>", SPF_DEFAULT, NULL);

 

最后，以一段简单的报时结束这篇文章：

 

SYSTEMTIME st;

GetLocalTime(&st);

 

ISpObjectToken * pSpObjectToken = NULL;

if (SUCCEEDED(SpFindBestToken(SPCAT_VOICES, L"language=804", NULL, &pSpObjectToken)))

{

       pSpVoice->SetVoice(pSpObjectToken);

 

       WCHAR wsz[128];

       swprintf(wsz,

              L"现在是<context ID = /"CHS_Date_ymdhm/">%.4d-%.02d-%.2d %.2d:%.02d</context>",

              st.wYear, st.wMonth, st.wDay, st.wHour, st.wMinute);

 

       pSpVoice->Speak(wsz, SPF_IS_XML, NULL);

 

       pSpObjectToken->Release();

}

 

 

http://blog.csdn.net/lkjx82/archive/2006/07/23/965120.aspx

现在语音技术应用十分广泛，本人也一直在从事这方面的工作，利用MS TTS开发产品，针对中文发音设置的问题，在这里阐述一下自己的实现方法。

void CIMEInfoAcquisitionDlg::SpeakVoice(LPCTSTR csVoice)
{
   //com初期化
    if (FAILED(CoInitialize(NULL)))
    {
        AfxMessageBox("Error to intiliaze COM");
        return;
    }
 
 //文字変換
 LPOLESTR lpw = new WCHAR [256];
    MultiByteToWideChar(CP_ACP,0,csVoice,-1,lpw,256);

 HRESULT hr = S_OK;
    if( SUCCEEDED( hr ) )
 {
  CComPtr<ISpObjectToken>        cpVoiceToken;
  CComPtr<IEnumSpObjectTokens>   cpEnum;
  CComPtr<ISpVoice>              cpVoice;
  ULONG                          ulCount = 0;

  // 建立

  hr = cpVoice.CoCreateInstance(CLSID_SpVoice);

  LPCSTR name;
    name = "Microsoft Simplified Chinese";

  CString s; s.Format("Name=%s", name);
  WCHAR wName[1024];
  MultiByteToWideChar(CP_ACP, 0, s, -1, wName, sizeof(wName));

  ISpObjectToken* pToken = NULL;
  if(SUCCEEDED(SpFindBestToken(SPCAT_VOICES, NULL, wName, &pToken))){
  cpVoice->SetVoice(pToken); 
  }
  //文读
     if (SUCCEEDED (hr))
  {
     hr = cpVoice->Speak( lpw, SPF_DEFAULT, NULL );
  }

 }

 //com的结束
    delete [] lpw;
    CoUninitialize();
 return;
}

以上利用ISpObjectToken接口实现了对中文的阅读，是不是很简单明了：）

 

http://blog.csdn.net/lkjx82/archive/2006/07/23/965120.aspx

 

http://www.pcdog.com/edu/vc/2005/12/c057683.html

http://topic.csdn.net/t/20050303/17/3822247.html