/// <summary>
/// Creates a new speech recognition engine
/// </summary>
/// <returns>A new speech recognition engine object.</returns>
private SpeechRecognitionEngine CreateSpeechRecognitionEngine()
{
// Create the recognizer
var recognizer = SystemSpeech.CreateSpeechRecognitionEngine(this.Configuration.Language, this.Configuration.Grammars);
// Attach event handlers for speech recognition events
recognizer.SpeechDetected += this.OnSpeechDetected;
recognizer.SpeechHypothesized += this.OnSpeechHypothesized;
recognizer.SpeechRecognized += this.OnSpeechRecognized;
recognizer.SpeechRecognitionRejected += this.OnSpeechRecognitionRejected;
recognizer.AudioSignalProblemOccurred += this.OnAudioSignalProblemOccurred;
recognizer.AudioStateChanged += this.OnAudioStateChanged;
recognizer.RecognizeCompleted += this.OnRecognizeCompleted;
recognizer.RecognizerUpdateReached += this.OnRecognizerUpdateReached;
recognizer.AudioLevelUpdated += this.OnAudioLevelUpdated;
recognizer.EmulateRecognizeCompleted += this.OnEmulateRecognizeCompleted;
recognizer.LoadGrammarCompleted += this.OnLoadGrammarCompleted;
// Create the format info from the configuration input format
SpeechAudioFormatInfo formatInfo = new SpeechAudioFormatInfo(
(EncodingFormat)this.Configuration.InputFormat.FormatTag,
(int)this.Configuration.InputFormat.SamplesPerSec,
this.Configuration.InputFormat.BitsPerSample,
this.Configuration.InputFormat.Channels,
(int)this.Configuration.InputFormat.AvgBytesPerSec,
this.Configuration.InputFormat.BlockAlign,
(this.Configuration.InputFormat is WaveFormatEx) ? ((WaveFormatEx)this.Configuration.InputFormat).ExtraInfo : null);
// Specify the input stream and audio format
recognizer.SetInputToAudioStream(this.inputAudioStream, formatInfo);
return(recognizer);
}
/// <summary>
/// Creates a new speech recognition engine.
/// </summary>
/// <returns>A new speech recognition engine object.</returns>
private SpeechRecognitionEngine CreateSpeechRecognitionEngine()
{
// Create speech recognition engine
var recognizer = SystemSpeech.CreateSpeechRecognitionEngine(this.Configuration.Language, this.Configuration.Grammars);
// Attach event handlers for speech recognition events
recognizer.AudioStateChanged += this.OnAudioStateChanged;
recognizer.RecognizeCompleted += this.OnRecognizeCompleted;
// Create the format info from the configuration input format
SpeechAudioFormatInfo formatInfo = new SpeechAudioFormatInfo(
(EncodingFormat)this.Configuration.InputFormat.FormatTag,
(int)this.Configuration.InputFormat.SamplesPerSec,
this.Configuration.InputFormat.BitsPerSample,
this.Configuration.InputFormat.Channels,
(int)this.Configuration.InputFormat.AvgBytesPerSec,
this.Configuration.InputFormat.BlockAlign,
(this.Configuration.InputFormat is WaveFormatEx) ? ((WaveFormatEx)this.Configuration.InputFormat).ExtraInfo : null);
// Specify the input stream and audio format
recognizer.SetInputToAudioStream(this.inputAudioStream, formatInfo);
// Set the speech recognition engine parameters
recognizer.InitialSilenceTimeout = TimeSpan.FromMilliseconds(this.Configuration.InitialSilenceTimeoutMs);
recognizer.BabbleTimeout = TimeSpan.FromMilliseconds(this.Configuration.BabbleTimeoutMs);
recognizer.EndSilenceTimeout = TimeSpan.FromMilliseconds(this.Configuration.EndSilenceTimeoutMs);
recognizer.EndSilenceTimeoutAmbiguous = TimeSpan.FromMilliseconds(this.Configuration.EndSilenceTimeoutAmbiguousMs);
return(recognizer);
}
private SpeechRecognitionEngine CreateSpeechRecognitionEngine()
{
// Create the recognizer
var recognizer = SystemSpeech.CreateSpeechRecognitionEngine(this.Configuration.Language, this.Configuration.Grammars);
// Attach event handlers for speech recognition events
recognizer.LoadGrammarCompleted += this.OnLoadGrammarCompleted;
return(recognizer);
}
/// <inheritdoc/>
protected override void Receive(string text, Envelope e)
{
if (string.IsNullOrWhiteSpace(text))
{
this.Out.Post(new IntentData(), e.OriginatingTime);
}
else
{
var result = this.speechRecognitionEngine.EmulateRecognize(text);
if (result != null && result.Semantics != null)
{
var intents = SystemSpeech.BuildIntentData(result.Semantics);
this.Out.Post(intents, e.OriginatingTime);
}
else
{
this.Out.Post(new IntentData(), e.OriginatingTime);
}
}
}
/// <summary>
/// Called when the final recognition result received.
/// </summary>
/// <param name="sender">The source of the event.</param>
/// <param name="e">An object that contains the event data.</param>
private void OnSpeechRecognized(object sender, SpeechRecognizedEventArgs e)
{
// Convention for intervals is to use the end time as the originating time so we add the duration as well.
DateTime originatingTime = this.streamStartTime + e.Result.Audio.AudioPosition + e.Result.Audio.Duration;
// Post the raw result from the underlying recognition engine
this.PostWithOriginatingTimeConsistencyCheck(this.SpeechRecognized, e, originatingTime);
if (e.Result.Alternates.Count > 0)
{
var result = this.BuildSpeechRecognitionResult(e.Result);
this.PostWithOriginatingTimeConsistencyCheck(this.Out, result, originatingTime);
if (e.Result.Semantics != null)
{
var intents = SystemSpeech.BuildIntentData(e.Result.Semantics);
this.PostWithOriginatingTimeConsistencyCheck(this.IntentData, intents, originatingTime);
}
}
}
