Share via

SpeechRecognitionConnector

  • Article

A SpeechRecognitionConnector instance can transfer audio data from a SpeechRecognitionEngine instance (in the Microsoft.Speech namespace) to an AudioVideoFlow instance that is attached to the SpeechRecognitionConnector instance.

The stream of audio data produced by a SpeechRecognitionConnector instance is compatible with that produced by the Microsoft.Speech namespace in Microsoft .NET Framework 3.0 and later. One such stream is provided every time Start is called on the SpeechRecognitionConnector. The stream contains the audio up to the point at which Stop is called on the connector. Calling Start again then generates a new stream for the new interval.

Before an application can call the Start method on a SpeechRecognitionConnector instance, an AudioVideoFlow instance must have been previously attached to the connector, and the State of the AudioVideoFlow instance must be Active. If both conditions are not met, Start throws an exception.

Example

The following procedure describes the steps involved in creating and using a SpeechRecognitionConnector instance.

To use SpeechRecognitionConnector

  1. Create an instance of this class.

    SpeechRecognitionConnector recoConn = new SpeechRecognitionConnector();

  2. Attach an AudioVideoFlow instance (assumed to have been previously constructed and in the Active state).

    recoConn.AttachFlow(AudioVideoFlow);

  3. Start the connector to initiate speech recognition.

    SpeechRecognitionStream recoStream = SpeechRecognitionConnector.Start();

    This step alerts the AudioVideoFlow to the existence of the connector and supplies the application with a stream of audio data that begins at roughly the time that Start was called. Start throws an exception if another receiving device is already active. The stream that is supplied is of type SpeechRecognitionStream, a type that inherits from Stream.

  4. When the application is finished with its speech recognition, it calls Stop.

    recoConn.Stop();

    The stream remains open-ended (that is, a call to Read will block when there are no media samples returned by the connector, rather than return "End of Stream") until the application calls Stop.

  5. The application owns the stream as soon as it is returned by Start, so when it is finished with the stream, it should call Dispose.

    recoStream.Dispose();

    This will return any buffers remaining in the stream to the SpeechRecognitionConnector connector.

  6. When the application is completely finished with the connector itself, it should call Dispose to free all unmanaged buffers.

    recoConn.Dispose();