none
Audio Level not updated? RRS feed

  • Question

  • I had the Kinect's USB audio microphone array volume set to the max level, and the sound control panel shows that there is audio input activity for mic input. However, I can't seem to receive any AudioLevelUpdated events when using the Microsoft.Speech.SpeechRecognitonEngine with KinectAudioSource's as input audio stream. Did someone figure this out, or is this a bug with the API?

    Tuesday, December 20, 2011 11:41 PM

Answers

  • At this time, this is an expected behavior of the speech runtime.  Currently, AudioLevelUpdated event is applicable for cases when using the default audio device (SpeechRecognitionEngine.SetInputToDefaultAudioDevice).

    KinectAudioSource only starts and feeds an audio stream to the recognizer via SetInputToAudioStream, so audio levels aren’t
    applied. Enabling this event for Kinect Audio is something we may consider in the future, but is not currently supported.

    Wednesday, April 4, 2012 7:21 PM

All replies

  • Please ensure that the audio is set to 3, not max.  3 will ensure that there is a 0 dB gain on the signal, which will get better recognition performance out of the recognizer. 

    I am seeing the same behavior and looking into it... Out of curiousity, why do you want the event?  Generally you can just let the recognizer run and handle the speech output rather than accessing the audio levels directly, so I'd love to understand your scenario a bit more.

    Tuesday, April 3, 2012 5:37 PM
  • At this time, this is an expected behavior of the speech runtime.  Currently, AudioLevelUpdated event is applicable for cases when using the default audio device (SpeechRecognitionEngine.SetInputToDefaultAudioDevice).

    KinectAudioSource only starts and feeds an audio stream to the recognizer via SetInputToAudioStream, so audio levels aren’t
    applied. Enabling this event for Kinect Audio is something we may consider in the future, but is not currently supported.

    Wednesday, April 4, 2012 7:21 PM
  • I'm trying to use the audio levels to display visual feedback on Kinect's mic input, this is a fundamental UX requirement. My PC contains more than one microphone input devices, which could be set to default by the User.  I certainly want the speech recognition engine to always analyze audio stream from Kinect's mic array, instead of the stereo mic from the PC.
    Thursday, July 26, 2012 4:45 AM