From a45d4b2d7db0f191ef54e8b7aef1eccf3d5c7534 Mon Sep 17 00:00:00 2001
From: David Rowe <david@ctrlaltstudio.com>
Date: Wed, 22 Oct 2014 10:17:29 -0700
Subject: [PATCH] Add first pass at Windows speech recognizer implementation

---
 interface/src/SpeechRecognizer.cpp | 135 +++++++++++++++++++++++++++--
 interface/src/SpeechRecognizer.h   |  21 +++++
 2 files changed, 151 insertions(+), 5 deletions(-)

diff --git a/interface/src/SpeechRecognizer.cpp b/interface/src/SpeechRecognizer.cpp
index 02f3ece866..2faf4a1033 100644
--- a/interface/src/SpeechRecognizer.cpp
+++ b/interface/src/SpeechRecognizer.cpp
@@ -10,20 +10,39 @@
 //
 
 #include <QtGlobal>
-
-#ifdef Q_OS_WIN
+#include <QDebug>
 
 #include "SpeechRecognizer.h"
 
+#ifdef Q_OS_WIN
+
 SpeechRecognizer::SpeechRecognizer() :
     QObject(),
     _enabled(false),
-    _commands() {
+    _commands(),
+    _comInitialized(false),
+    _speechRecognizer(NULL),
+    _speechRecognizerContext(NULL),
+    _speechRecognizerGrammar(NULL),
+    _commandRecognizedEvent(NULL),
+    _commandRecognizedNotifier(NULL) {
 
+    HRESULT hr = ::CoInitialize(NULL);
+
+    if (SUCCEEDED(hr)) {
+        _comInitialized = true;
+    }
 }
 
 SpeechRecognizer::~SpeechRecognizer() {
 
+    if (_comInitialized) {
+        ::CoUninitialize();
+    }
+
+    if (_speechRecognizerGrammar) {
+        _speechRecognizerGrammar.Release();
+    }
 }
 
 void SpeechRecognizer::handleCommandRecognized(const char* command) {
@@ -31,15 +50,53 @@ void SpeechRecognizer::handleCommandRecognized(const char* command) {
 }
 
 void SpeechRecognizer::setEnabled(bool enabled) {
-    if (enabled == _enabled) {
+    if (enabled == _enabled || !_comInitialized) {
         return;
     }
 
     _enabled = enabled;
+
     if (_enabled) {
 
-    } else {
+        HRESULT hr = S_OK;
 
+        // Create a shared recognizer.
+        if (SUCCEEDED(hr)) {
+            hr = _speechRecognizerContext.CoCreateInstance(CLSID_SpSharedRecoContext);
+        }
+
+        // Set up event notification mechanism.
+        if (SUCCEEDED(hr)) {
+            hr = _speechRecognizerContext->SetNotifyWin32Event();
+        }
+        if (SUCCEEDED(hr)) {
+            _commandRecognizedEvent = _speechRecognizerContext->GetNotifyEventHandle();
+            if (_commandRecognizedEvent == NULL) {
+                hr = S_FALSE;
+            }
+        }
+        if (SUCCEEDED(hr)) {
+            _commandRecognizedNotifier = new QWinEventNotifier(_commandRecognizedEvent);
+            connect(_commandRecognizedNotifier, SIGNAL(activated(HANDLE)), SLOT(notifyCommandRecognized(HANDLE)));
+        }
+        
+        // Set which events to be notified of.
+        if (SUCCEEDED(hr)) {
+            hr = _speechRecognizerContext->SetInterest(SPFEI(SPEI_RECOGNITION), SPFEI(SPEI_RECOGNITION));
+        }
+
+        // Create grammar and load commands.
+        if (SUCCEEDED(hr)) {
+            hr = _speechRecognizerContext->CreateGrammar(NULL, &_speechRecognizerGrammar);
+        }
+        if (SUCCEEDED(hr)) {
+            reloadCommands();
+        }
+        
+        _enabled = SUCCEEDED(hr);
+
+    } else {
+        _speechRecognizerContext.Release();
     }
 
     emit enabledUpdated(_enabled);
@@ -56,7 +113,75 @@ void SpeechRecognizer::removeCommand(const QString& command) {
 }
 
 void SpeechRecognizer::reloadCommands() {
+    if (!_enabled) {
+        return;
+    }
 
+    HRESULT hr = S_OK;
+
+    if (SUCCEEDED(hr)) {
+        hr = _speechRecognizerContext->Pause(NULL);
+    }
+
+    if (SUCCEEDED(hr)) {
+        WORD langId = MAKELANGID(LANG_NEUTRAL, SUBLANG_NEUTRAL);
+        hr = _speechRecognizerGrammar->ResetGrammar(langId);
+    }
+
+    DWORD ruleID = 0;
+    SPSTATEHANDLE initialState;
+    for (QSet<QString>::const_iterator iter = _commands.constBegin(); iter != _commands.constEnd(); iter++) {
+        ruleID += 1;
+
+        if (SUCCEEDED(hr)) {
+            hr = _speechRecognizerGrammar->GetRule(NULL, ruleID, SPRAF_TopLevel | SPRAF_Active | SPRAF_Dynamic, TRUE, &initialState);
+        }
+
+        if (SUCCEEDED(hr)) {
+            const std::wstring command = (*iter).toStdWString();  // DJRTODO: Better, shorter name
+            hr = _speechRecognizerGrammar->AddWordTransition(initialState, NULL, command.c_str(), L" ", SPWT_LEXICAL, 1.0, NULL);
+        }
+    }
+
+    if (SUCCEEDED(hr)) {
+        hr = _speechRecognizerGrammar->Commit(0);
+    }
+
+    if (SUCCEEDED(hr)) {
+        hr = _speechRecognizerContext->Resume(NULL);
+    }
+
+    if (SUCCEEDED(hr)) {
+        hr = _speechRecognizerGrammar->SetRuleState(NULL, NULL, SPRS_ACTIVE);
+    }
+
+    if (FAILED(hr)) {
+        qDebug() << "ERROR: Didn't successfully reload commands";
+    }
+}
+
+void SpeechRecognizer::notifyCommandRecognized(HANDLE handle) {
+
+    SPEVENT eventItem;
+    memset(&eventItem, 0, sizeof(SPEVENT));
+    HRESULT hr = _speechRecognizerContext->GetEvents(1, &eventItem, NULL);
+
+    if (SUCCEEDED(hr)) {
+        if (eventItem.eEventId == SPEI_RECOGNITION && eventItem.elParamType == SPET_LPARAM_IS_OBJECT) {
+            ISpRecoResult* recognitionResult = reinterpret_cast<ISpRecoResult*>(eventItem.lParam);
+            wchar_t* pText;
+
+            hr = recognitionResult->GetText(SP_GETWHOLEPHRASE, SP_GETWHOLEPHRASE, FALSE, &pText, NULL);
+
+            if (SUCCEEDED(hr)) {
+                QString text = QString::fromWCharArray(pText);
+                handleCommandRecognized(text.toStdString().c_str());
+                ::CoTaskMemFree(pText);
+            }
+
+            recognitionResult->Release();
+        }
+    }
 }
 
 #endif // Q_OS_WIN
\ No newline at end of file
diff --git a/interface/src/SpeechRecognizer.h b/interface/src/SpeechRecognizer.h
index edd4abe1d6..b343326de3 100644
--- a/interface/src/SpeechRecognizer.h
+++ b/interface/src/SpeechRecognizer.h
@@ -16,6 +16,13 @@
 #include <QSet>
 #include <QString>
 
+#ifdef Q_OS_WIN
+#include <QWinEventNotifier>
+
+#include <atlbase.h>
+#include <sapi.h>
+#endif
+
 class SpeechRecognizer : public QObject {
     Q_OBJECT
 public:
@@ -40,8 +47,22 @@ protected:
 private:
     bool _enabled;
     QSet<QString> _commands;
+#if defined(Q_OS_MAC)
     void* _speechRecognizerDelegate;
     void* _speechRecognizer;
+#elif defined(Q_OS_WIN)
+    bool _comInitialized;
+    CComPtr<ISpRecognizer> _speechRecognizer;
+    CComPtr<ISpRecoContext> _speechRecognizerContext;
+    CComPtr<ISpRecoGrammar> _speechRecognizerGrammar;
+    HANDLE _commandRecognizedEvent;
+    QWinEventNotifier* _commandRecognizedNotifier;
+#endif
+
+#if defined(Q_OS_WIN)
+private slots:
+    void notifyCommandRecognized(HANDLE handle);
+#endif
 };
 
 #endif // hifi_SpeechRecognizer_h