JunkBox_Win_Lib: WinNiLib/NiSpeech.cpp Source File

00001 
00012 #include  "WinBaseLib.h"
00013 #include  "NiSpeech.h"
00014 
00015 
00016 
00017 #ifdef  ENABLE_NI_SPEECH
00018 
00019 
00020 using namespace jbxl;
00021 using namespace jbxwl;
00022 
00023 
00025 // Face Tracker クラス
00026 
00027 CNiSpeech::CNiSpeech(IStream* stream, WAVEFORMATEX* format)
00028 {
00029     m_pSpeechStream     = NULL;
00030     m_pSpeechRecognizer = NULL;
00031     m_pSpeechContext    = NULL;
00032     m_pSpeechGrammar    = NULL;
00033 
00034     m_hSpeechEvent      = NULL;
00035     m_hStopEvent        = NULL;
00036     m_speechThread      = NULL;
00037 
00038     m_confidence        = 0.1;
00039 
00040     if (stream!=NULL && format!=NULL) init(stream, format);
00041 }
00042 
00043 
00044 
00045 BOOL  CNiSpeech::init(IStream* stream, WAVEFORMATEX* format)
00046 {
00047     DEBUG_INFO("CNiSpeech::init(): START\n");
00048     free();
00049 
00050     if (stream==NULL || format==NULL) return FALSE;
00051 
00052     HRESULT hr = CoCreateInstance(CLSID_SpStream, NULL, CLSCTX_INPROC_SERVER, __uuidof(ISpStream), (void**)&m_pSpeechStream);
00053     if (FAILED(hr)) {
00054         DEBUG_INFO("CNiSpeech::init(): ERR 1\n");
00055         return FALSE;
00056     }
00057 
00058     hr = m_pSpeechStream->SetBaseStream(stream, SPDFID_WaveFormatEx, format);
00059     if (FAILED(hr)) {
00060         DEBUG_INFO("CNiSpeech::init(): ERR 2\n");
00061         releaseNull(m_pSpeechStream);
00062         return FALSE;
00063     }
00064 
00065     DEBUG_INFO("CNiSpeech::init(): END\n");
00066     return TRUE;
00067 }
00068 
00069 
00070 
00071 void  CNiSpeech::free(void)
00072 {
00073     DEBUG_INFO("CNiSpeech::free(): START\n");
00074     releaseNull(m_pSpeechGrammar);
00075     releaseNull(m_pSpeechContext);
00076     releaseNull(m_pSpeechRecognizer);
00077     //releaseNull(m_pSpeechStream); // sometimes crash 
00078     DEBUG_INFO("CNiSpeech::free(): END\n");
00079 }
00080 
00081 
00082 
00083 BOOL  CNiSpeech::create(LPCTSTR lang)
00084 {
00085     DEBUG_INFO("CNiSpeech::create(): START\n");
00086     ISpObjectToken *pEngineToken = NULL;
00087 
00088     releaseNull(m_pSpeechContext);
00089     releaseNull(m_pSpeechRecognizer);
00090    
00091     HRESULT hr = CoCreateInstance(CLSID_SpInprocRecognizer, NULL, CLSCTX_INPROC_SERVER, __uuidof(ISpRecognizer), (void**)&m_pSpeechRecognizer);
00092     if (FAILED(hr)) {
00093         DEBUG_INFO("CNiSpeech::create(): ERR 1\n");
00094         return FALSE;
00095     }
00096 
00097     m_pSpeechRecognizer->SetInput(m_pSpeechStream, FALSE);
00098     hr = SpFindBestToken(SPCAT_RECOGNIZERS, lang, NULL, &pEngineToken);
00099 
00100     if (SUCCEEDED(hr)) {
00101         m_pSpeechRecognizer->SetRecognizer(pEngineToken);
00102         hr = m_pSpeechRecognizer->CreateRecoContext(&m_pSpeechContext);
00103         releaseNull(pEngineToken);
00104     }
00105 
00106     if (FAILED(hr)) {
00107         DEBUG_INFO("CNiSpeech::create(): ERR 2\n");
00108         releaseNull(m_pSpeechContext);
00109         releaseNull(m_pSpeechRecognizer);
00110         releaseNull(m_pSpeechStream);
00111         return FALSE;
00112     }
00113 
00114     DEBUG_INFO("CNiSpeech::create(): END\n");
00115     return TRUE;
00116 }
00117 
00118 
00119 
00120 BOOL  CNiSpeech::load(LPCTSTR file)
00121 {
00122     DEBUG_INFO("CNiSpeech::load(): START\n");
00123 
00124     if (m_pSpeechContext==NULL) {
00125         DEBUG_INFO("CNiSpeech::load(): SpeechContext ERR\n");
00126         return FALSE;
00127     }
00128 
00129     releaseNull(m_pSpeechGrammar);
00130 
00131     HRESULT hr = m_pSpeechContext->CreateGrammar(1, &m_pSpeechGrammar);
00132 
00133     if (SUCCEEDED(hr)) hr = m_pSpeechGrammar->LoadCmdFromFile(file, SPLO_STATIC);
00134     if (FAILED(hr)) {
00135         DEBUG_INFO("CNiSpeech::load: ERR 2\n");
00136         releaseNull(m_pSpeechGrammar);
00137         return FALSE;
00138     }
00139 
00140     DEBUG_INFO("CNiSpeech::load(): END\n");
00141     return TRUE;
00142 }
00143 
00144 
00145 
00146 BOOL  CNiSpeech::start(double confidence=0.1)
00147 {
00148     if (m_pSpeechGrammar==NULL || m_pSpeechRecognizer==NULL || m_pSpeechContext==NULL) return FALSE;
00149     DEBUG_INFO("CNiSpeech::start(): START\n");
00150 
00151 //  stop();     // if working then stop
00152 
00153     m_confidence = confidence;
00154     //
00155     m_pSpeechGrammar->SetRuleState(NULL, NULL, SPRS_ACTIVE);
00156     m_pSpeechRecognizer->SetRecoState(SPRST_ACTIVE_ALWAYS);
00157     DEBUG_INFO("CNiSpeech::start(): SetRecoState OK\n");
00158     ::DisPatcher();
00159     ::Sleep(100);
00160 
00161     m_pSpeechContext->SetInterest(SPFEI(SPEI_RECOGNITION), SPFEI(SPEI_RECOGNITION));
00162     HRESULT hr = m_pSpeechContext->Resume(0);
00163     if (FAILED(hr)) {
00164         DEBUG_INFO("CNiSpeech::start(): Resume ERR\n");
00165         free();
00166         return FALSE;
00167     }
00168     DEBUG_INFO("CNiSpeech::start(): Resume OK\n");
00169 
00170     m_hStopEvent   = CreateEvent(NULL, FALSE, FALSE, NULL);
00171     m_hSpeechEvent = m_pSpeechContext->GetNotifyEventHandle();
00172     //
00173     m_speechThread = AfxBeginThread(speechThread, (LPVOID)this, THREAD_PRIORITY_NORMAL, 0, CREATE_SUSPENDED);
00174     m_speechThread->m_bAutoDelete = FALSE;
00175     m_speechThread->ResumeThread();
00176     ::Sleep(10);
00177 
00178     DEBUG_INFO("CNiSpeech::start(): END");
00179     return TRUE;
00180 }
00181 
00182 
00183 
00184 void  CNiSpeech::stop(void)
00185 {
00186     DEBUG_INFO("CNiSpeech::stop(): START\n");
00187 
00188     if (m_hStopEvent!=NULL) {
00189         SetEvent(m_hStopEvent);
00190         //
00191         if (m_speechThread!=NULL) {
00192             WaitForSingleObject(m_speechThread->m_hThread, INFINITE);
00193             CloseHandle(m_speechThread->m_hThread);
00194             delete m_speechThread;  // as m_bAutoDelete==FALSE 
00195             m_speechThread = NULL;
00196             //
00197             //m_pSpeechContext->Pause(0);
00198             DEBUG_INFO("CNiSpeech::stop(): Pause OK\n");
00199        }
00200         CloseHandle(m_hStopEvent);
00201         m_hStopEvent = NULL;
00202     }
00203 
00204     //
00205     if (m_hSpeechEvent!=NULL) {
00206         CloseHandle(m_hSpeechEvent);
00207         m_hSpeechEvent = NULL;
00208     }
00209 
00210     DEBUG_INFO("CNiSpeech::stop(): END\n");
00211     return;
00212 }
00213 
00214 
00215 
00216 void  CNiSpeech::process(void)
00217 {
00218     SPEVENT curEvent;
00219     ULONG   fetched = 0;
00220 
00221     m_pSpeechContext->GetEvents(1, &curEvent, &fetched);
00222 
00223     while (fetched>0) {
00224         if (curEvent.eEventId==SPEI_RECOGNITION && curEvent.elParamType==SPET_LPARAM_IS_OBJECT) {
00225             //
00226             ISpRecoResult* result = reinterpret_cast<ISpRecoResult*>(curEvent.lParam);
00227             SPPHRASE* pPhrase = NULL;
00228             result->GetPhrase(&pPhrase);
00229             //
00230             if (pPhrase!=NULL) {
00231                 if (pPhrase->pProperties!=NULL && pPhrase->pProperties->pFirstChild!=NULL) {
00232                     const SPPHRASEPROPERTY* pSemanticTag = pPhrase->pProperties->pFirstChild;
00233                     //
00234                     map2action(pSemanticTag->pszValue, pSemanticTag->SREngineConfidence);
00235                 }
00236                 ::CoTaskMemFree(pPhrase);
00237             }
00238         }
00239 
00240         m_pSpeechContext->GetEvents(1, &curEvent, &fetched);
00241     }
00242 
00243     return;
00244 }
00245 
00246 
00247 
00248 void  CNiSpeech::setConfidence(double confd)
00249 {
00250     if (confd>1.0)      confd = 1.0;
00251     else if (confd<0.0) confd = 0.0;
00252 
00253     m_confidence = confd;
00254 }
00255 
00256 
00257 
00258 
00260 // スレッド
00261 
00262 UINT  CNiSpeech::speechThread(LPVOID pParam)
00263 {
00264     CNiSpeech* pthis = (CNiSpeech*)pParam;
00265     return pthis->speechThread();
00266 }
00267 
00268 
00269 
00270 UINT  CNiSpeech::speechThread(void)
00271 {
00272     DEBUG_INFO("CNiSpeech::speechThread(): START\n");
00273 
00274     bool bContinue = true;
00275     
00276     while(bContinue) {
00277         //
00278         if (WaitForSingleObject(m_hStopEvent, 0)==WAIT_OBJECT_0) {
00279             bContinue = false;
00280             continue;
00281         }
00282         
00283         if (WaitForSingleObject(m_hSpeechEvent, 0)==WAIT_OBJECT_0) {
00284             process();
00285         }
00286 
00287         Sleep(10);
00288     }
00289 
00290     DEBUG_INFO("CNiSpeech::speechThread(): END\n");
00291     return TRUE;
00292 }
00293 
00294 
00295 
00296 void  CNiSpeech::map2action(LPCTSTR tag, double confd)
00297 {
00298     WORD param;
00299 
00300     if (confd>=m_confidence) {
00301         param = TRUE;
00302         SendWinMessage(JBXWL_WM_SPEECH_EVENT, (WPARAM)&param, (LPARAM)tag);
00303     }
00304     else {
00305         param = FALSE;
00306         SendWinMessage(JBXWL_WM_SPEECH_EVENT, (WPARAM)&param, (LPARAM)_T("N/A"));
00307     }
00308 }
00309 
00310 
00311 #endif      // ENABLE_NI_SPEECH
00312