none
AudioBeam Stream for speech recognition does not take into account the Beam Angle RRS feed

  • Question

  • I'm trying to create an app that gets the audio stream from the AudioBeam to provide it to the speech recognition engine.

    I have that working but I have a couple issues with this.

    1) The speech recognition engine is waiting for a silence to try to come up with a result (set of words) of what's being said. Although I'm handling this correctly right now. If the audio stream does not have a silence then I dont get any "streamed" results. Its a problem when I have ambient noise because there is no silence for the speech recognizer to analyze the phrase.

     - I wonder if there is a way to control the audio stream provided to the speech recognizer engine, or if there is a property to reduce the silence needed for it to recognize the words.

    2) another major issue is that I want only words that are coming from in front of the kinect to be recognized by the engine. Right now, the audio stream seems to be taking audio from all directions. Any solution for this issue?

    - I thought that if the problem is that the raw audio stream is not filtering by the beam angle and confidence. I was trying to handle the frame recieved event of the audio reader in which I can access the subframe for angle and audio body correlation and trying to insert such subframe into a stream to pass to the speech recognizer. but, maybe I seem to have a problem because I never get the Complete recognition event.

    Thanks in advance.

    Friday, March 24, 2017 1:31 AM