SpeechRecognition: Add support for resetting engine adaptation state

- Add QSpeechRecognitionEngine::resetAdaptationState()
- Add QSpeechRecognitionPluginEngine::resetAdaptationState()
- Implement the feature for PocketSphinx

Change-Id: I68411064fb99014558be40f9187a453fcc546799
Reviewed-by: Andrew Knight <andrew.knight@intopalo.com>

11 files changed, 64 insertions, 0 deletions

diff --git a/src/asr/qspeechrecognition.cpp b/src/asr/qspeechrecognition.cpp
index ad7c39f..2d3a53d 100755
--- a/src/asr/qspeechrecognition.cpp
+++ b/src/asr/qspeechrecognition.cpp
@@ -343,6 +343,7 @@ QSpeechRecognitionEngine *QSpeechRecognition::createEngine(const QString &name,
         return 0; // Already exists
     QSpeechRecognitionEngineImpl *engine = new QSpeechRecognitionEngineImpl(name, this);
     connect(engine, &QSpeechRecognitionEngineImpl::requestSetParameter, d->m_managerInterface, &QSpeechRecognitionManagerInterface::onSetEngineParameter);
+    connect(engine, &QSpeechRecognitionEngineImpl::requestResetAdaptationState, d->m_managerInterface, &QSpeechRecognitionManagerInterface::onResetEngineAdaptationState);
     d->m_engines.insert(name, engine);
     emit d->m_managerInterface->createEngine(name, providerName, parameters);
     return engine;
@@ -657,6 +658,7 @@ QSpeechRecognitionPrivate::QSpeechRecognitionPrivate():
     QObject::connect(m_managerInterface, &QSpeechRecognitionManagerInterface::reset, m_manager, &QSpeechRecognitionManager::reset, Qt::QueuedConnection);
     QObject::connect(m_managerInterface, &QSpeechRecognitionManagerInterface::dispatchMessage, m_manager, &QSpeechRecognitionManager::dispatchMessage, Qt::QueuedConnection);
     QObject::connect(m_managerInterface, &QSpeechRecognitionManagerInterface::setEngineParameter, m_manager, &QSpeechRecognitionManager::setEngineParameter, Qt::QueuedConnection);
+    QObject::connect(m_managerInterface, &QSpeechRecognitionManagerInterface::resetEngineAdaptationState, m_manager, &QSpeechRecognitionManager::resetEngineAdaptationState, Qt::QueuedConnection);
 
     m_managerThread->start();
 }
diff --git a/src/asr/qspeechrecognition_p.h b/src/asr/qspeechrecognition_p.h
index 60f01ab..fad6125 100755
--- a/src/asr/qspeechrecognition_p.h
+++ b/src/asr/qspeechrecognition_p.h
@@ -166,6 +166,12 @@ public slots:
         if (engine)
             emit setEngineParameter(engine->name(), key, value);
     }
+    void onResetEngineAdaptationState()
+    {
+        QSpeechRecognitionEngineImpl *engine = qobject_cast<QSpeechRecognitionEngineImpl *>(QObject::sender());
+        if (engine)
+            emit resetEngineAdaptationState(engine->name());
+    }
 
 signals:
     void setSession(int session);
@@ -181,6 +187,7 @@ signals:
     void reset();
     void dispatchMessage(const QString &engineName, const QString &message, const QVariantMap &parameters);
     void setEngineParameter(const QString &engineName, const QString &key, const QVariant &value);
+    void resetEngineAdaptationState(const QString &engineName);
 
 private:
     QSpeechRecognitionPrivate *m_speech;
diff --git a/src/asr/qspeechrecognitionengine.cpp b/src/asr/qspeechrecognitionengine.cpp
index 5ecdce6..f7d822c 100755
--- a/src/asr/qspeechrecognitionengine.cpp
+++ b/src/asr/qspeechrecognitionengine.cpp
@@ -250,6 +250,17 @@ Q_DEFINE_ASR_ENGINE_PARAMETER(AudioInputDevices);
   \sa supportedParameters()
 */
 
+/*!
+  \fn void QSpeechRecognitionEngine::resetAdaptationState()
+
+  Reset engine adaptation state to its initial value.
+
+  If the engine has been adapted to user speech or audio path features,
+  the adaptation state is returned where it was after the first initialization,
+  before any audio data was fed to it. Any files that are used for storing the
+  adaptation data are deleted.
+*/
+
 QSpeechRecognitionEngine::QSpeechRecognitionEngine(QObject *parent):
     QObject(parent)
 {
@@ -291,6 +302,11 @@ QVariant QSpeechRecognitionEngineImpl::parameter(const QString &key) const
     return m_parameters.value(key);
 }
 
+void QSpeechRecognitionEngineImpl::resetAdaptationState()
+{
+    emit requestResetAdaptationState();
+}
+
 bool QSpeechRecognitionEngineImpl::isCreated()
 {
     return m_created;
diff --git a/src/asr/qspeechrecognitionengine.h b/src/asr/qspeechrecognitionengine.h
index 7681f17..afea163 100644
--- a/src/asr/qspeechrecognitionengine.h
+++ b/src/asr/qspeechrecognitionengine.h
@@ -61,6 +61,7 @@ public:
     Q_INVOKABLE virtual bool setParameter(const QString &key, const QVariant &value) = 0;
     Q_INVOKABLE virtual QVariant parameter(const QString &key) const = 0;
     Q_INVOKABLE virtual QList<QString> supportedParameters() const = 0;
+    Q_INVOKABLE virtual void resetAdaptationState() = 0;
     virtual bool isCreated() = 0;
 
     // Common engine parameter keys:
diff --git a/src/asr/qspeechrecognitionengine_p.h b/src/asr/qspeechrecognitionengine_p.h
index 7944865..8a23837 100755
--- a/src/asr/qspeechrecognitionengine_p.h
+++ b/src/asr/qspeechrecognitionengine_p.h
@@ -56,6 +56,7 @@ public:
     bool setParameter(const QString &key, const QVariant &value);
     QList<QString> supportedParameters() const;
     QVariant parameter(const QString &key) const;
+    void resetAdaptationState();
     bool isCreated();
     // Internal API:
     void setCreated(bool created);
@@ -65,6 +66,7 @@ public:
     QVariantMap m_parameters;
 signals:
     void requestSetParameter(const QString &key, const QVariant &value);
+    void requestResetAdaptationState();
 };
 
 QT_END_NAMESPACE
diff --git a/src/asr/qspeechrecognitionmanager.cpp b/src/asr/qspeechrecognitionmanager.cpp
index f3738db..687bafe 100755
--- a/src/asr/qspeechrecognitionmanager.cpp
+++ b/src/asr/qspeechrecognitionmanager.cpp
@@ -326,6 +326,12 @@ void QSpeechRecognitionManager::setEngineParameter(const QString &engineName, co
     }
 }
 
+void QSpeechRecognitionManager::resetEngineAdaptationState(const QString &engineName)
+{
+    if (QSpeechRecognitionPluginEngine *engine = m_engines.value(engineName, 0))
+        engine->resetAdaptationState();
+}
+
 void QSpeechRecognitionManager::onProcess()
 {
     QSet<QSpeechRecognitionPluginEngine*> engines = m_enginesToProcess;
diff --git a/src/asr/qspeechrecognitionmanager_p.h b/src/asr/qspeechrecognitionmanager_p.h
index 79f561d..86b2b7c 100755
--- a/src/asr/qspeechrecognitionmanager_p.h
+++ b/src/asr/qspeechrecognitionmanager_p.h
@@ -99,6 +99,7 @@ public slots:
     void reset();
     void dispatchMessage(const QString &engineName, const QString &message, const QVariantMap &parameters);
     void setEngineParameter(const QString &engineName, const QString &key, const QVariant &value);
+    void resetEngineAdaptationState(const QString &engineName);
 
 private slots:
     void onProcess();
diff --git a/src/asr/qspeechrecognitionpluginengine.cpp b/src/asr/qspeechrecognitionpluginengine.cpp
index 5cdb30b..7bd1f85 100755
--- a/src/asr/qspeechrecognitionpluginengine.cpp
+++ b/src/asr/qspeechrecognitionpluginengine.cpp
@@ -161,6 +161,13 @@ QT_BEGIN_NAMESPACE
   if \l stopListening() was called before this method.
 */
 
+/*! \fn void QSpeechRecognitionPluginEngine::resetAdaptationState()
+
+  Reset engine adaptation state to its initial value.
+
+  See QSpeechRecognitionEngine::resetAdaptationState().
+*/
+
 /*! \fn bool QSpeechRecognitionPluginEngine::process()
 
   Performs any run-time tasks the engine needs to do, like audio processing.
diff --git a/src/asr/qspeechrecognitionpluginengine.h b/src/asr/qspeechrecognitionpluginengine.h
index c7dba78..1a90719 100644
--- a/src/asr/qspeechrecognitionpluginengine.h
+++ b/src/asr/qspeechrecognitionpluginengine.h
@@ -71,6 +71,7 @@ public:
     virtual void abortListening() = 0;
     virtual void unmute(qint64 timestamp) = 0;
     virtual void reset() = 0;
+    virtual void resetAdaptationState() = 0;
     virtual bool process() = 0;
 
 Q_SIGNALS:
diff --git a/src/plugins/asr/pocketsphinx/qspeechrecognitionengine_pocketsphinx.cpp b/src/plugins/asr/pocketsphinx/qspeechrecognitionengine_pocketsphinx.cpp
index 8f6a799..7320b9b 100755
--- a/src/plugins/asr/pocketsphinx/qspeechrecognitionengine_pocketsphinx.cpp
+++ b/src/plugins/asr/pocketsphinx/qspeechrecognitionengine_pocketsphinx.cpp
@@ -88,6 +88,7 @@ QSpeechRecognitionEnginePocketSphinx::QSpeechRecognitionEnginePocketSphinx(const
         m_debugAudioFile(0),
         m_sessionStarted(false),
         m_cmnVec(0),
+        m_initialCmnVec(0),
         m_cmnSize(0)
 {
     const QVariantMap &engineParams = QSpeechRecognitionPluginEngine::parameters();
@@ -108,6 +109,7 @@ QSpeechRecognitionEnginePocketSphinx::~QSpeechRecognitionEnginePocketSphinx()
     disconnect(&m_inputFileDecoder);
     ps_free(m_decoder);
     delete[] m_cmnVec;
+    delete[] m_initialCmnVec;
 }
 
 bool QSpeechRecognitionEnginePocketSphinx::init(QString *errorString)
@@ -166,6 +168,8 @@ bool QSpeechRecognitionEnginePocketSphinx::init(QString *errorString)
     feat_t *feat = ps_get_feat(m_decoder);
     m_cmnSize = feat_cepsize(feat);
     m_cmnVec = new mfcc_t[m_cmnSize];
+    m_initialCmnVec = new mfcc_t[m_cmnSize];
+    cmn_prior_get(feat->cmn_struct, m_initialCmnVec);
     m_cmnFilePath = dataDirectory().absoluteFilePath(QLatin1String("pocketsphinx_")
                                                      + name() + QLatin1String("_cmn"));
     // Attempt to load adapted cepstrum means from the data file. The default values are not
@@ -375,6 +379,11 @@ void QSpeechRecognitionEnginePocketSphinx::reset()
     m_grammar = 0;
 }
 
+void QSpeechRecognitionEnginePocketSphinx::resetAdaptationState()
+{
+    resetCmn();
+}
+
 bool QSpeechRecognitionEnginePocketSphinx::process()
 {
     if (m_sessionStarted && !m_muted)
@@ -545,4 +554,13 @@ void QSpeechRecognitionEnginePocketSphinx::loadCmn()
     dataFile.close();
 }
 
+// Reset cepstrum means to their default values
+void QSpeechRecognitionEnginePocketSphinx::resetCmn()
+{
+    feat_t *feat = ps_get_feat(m_decoder);
+    cmn_prior_set(feat->cmn_struct, m_initialCmnVec);
+    QFile dataFile(m_cmnFilePath);
+    dataFile.remove();
+}
+
 QT_END_NAMESPACE
diff --git a/src/plugins/asr/pocketsphinx/qspeechrecognitionengine_pocketsphinx.h b/src/plugins/asr/pocketsphinx/qspeechrecognitionengine_pocketsphinx.h
index 899130c..eb13998 100755
--- a/src/plugins/asr/pocketsphinx/qspeechrecognitionengine_pocketsphinx.h
+++ b/src/plugins/asr/pocketsphinx/qspeechrecognitionengine_pocketsphinx.h
@@ -67,6 +67,7 @@ public:
     void abortListening();
     void unmute(qint64 timestamp);
     void reset();
+    void resetAdaptationState();
     bool process();
 
     // Internal API:
@@ -89,6 +90,7 @@ private:
     void processAudio(const void *data, size_t dataSize);
     void storeCmn();
     void loadCmn();
+    void resetCmn();
     int m_session;
     bool m_muted;
     ps_decoder_t *m_decoder;
@@ -104,6 +106,7 @@ private:
     bool m_sessionStarted;
     QString m_cmnFilePath;
     mfcc_t *m_cmnVec;
+    mfcc_t *m_initialCmnVec;
     int m_cmnSize;
 };