Skip to main content

Speech Recognition

All speech recognition functionality is handled through the SpeechRecogniser interface which can be accessed through the recogniser property on the main extension: Speech.instance.recogniser.

Start listening​

To process audio from the microphone call the startListening() method. This will activate an audio recording session and start passing the audio through the speech recogniser.

This process may dispatch the following events:

  • SpeechRecogniserEvent.RESULT: When the recogniser decodes audio to text;
  • SpeechRecogniserEvent.ERROR: If an error occurs;
Speech.instance.recogniser.addEventListener( SpeechRecogniserEvent.RESULT, resultHandler );
Speech.instance.recogniser.addEventListener( SpeechRecogniserEvent.ERROR, errorHandler );
Speech.instance.recogniser.startListening();

function resultHandler( event:SpeechRecogniserEvent ):void
{
    trace( "result: " + event.result.formattedString );
}

function errorHandler( event:SpeechRecogniserEvent ):void
{
    trace( "errorHandler [" + event.errorCode + "]:" + event.error )
}

startListening() will return a Boolean value, indicating whether the process was started successfully. It may return false if recognition is not supported or there was an error creating the audio session. (An error event may also be dispatched depending on the error).

On a SpeechRecogniserEvent.RESULT event there is a result property which is an instance of the SpeechRecogniserResult class. This class contains a formattedString representing all the detected speech in the audio. It also contains a segments array representing each of the individual detected words.

On a SpeechRecogniserEvent.ERROR there will be an errorCode that can be identified from the definitions in the SpeechRecogniserError class.

Options​

Speech recognition has a few options you can set via an instance of the SpeechRecogniserOptions class passed as the parameter to the startListening() method.

var options:SpeechRecogniserOptions = new SpeechRecogniserOptions();

Speech.instance.recogniser.startListening( options );

Using the options you can control whether you prefer "on device" or "network" recognition:

var options:SpeechRecogniserOptions = new SpeechRecogniserOptions()
    .setPreferOnDevice( false );

The default prefers on device, so if you want to use network enabled speech recognition (generally more accurate) you should disable this option as above.

Listening​

You can check the isListening flag at any point to determine if the extension is currently actively listening for speech.

if (Speech.instance.recogniser.isListening)
{
    // extension is listening
}

Stop listening​

Once you have completed your recognition process you must call stopListening() to end the audio session and stop the speech recognition task.

Speech.instance.recogniser.stopListening();

You can remove your event listeners at this point.