none
Kinect and Dictation RRS feed

  • Question

  • Dear all,

    I've tried for days only to find out that Microsoft.Speech.Recognition namespace does not support DictationGrammar.

    So, here's conflict:

    1. Dictation support seems to have been omitted from Microsoft.Speech.Recognition and is only present in System.Speech.Recognition

    2. Kinect only seems to work with Microsoft.Speech.Recognition, not System.Speech.Recognition

     

    Is there a way to use System.Speech.Recognition with Kinect?

     

     

    ----------------------------------------------

    [Update]

    I am temporarily using System.Speech.Recognition and using the default recognizer (MS Speech Recognition Engine 8.0 - US, FirstOrDefault()).

    I still want to use Kinect Speech Recognition Engine.

    Now that I am using Windows Speech Recognition Engine instead of Kinect's, is there any way to SetRecoProfile without using SpeechLib?

    Tuesday, June 21, 2011 1:08 AM

Answers

  • MuGMaNia's impression is correct. It is possible to support dictation scenarios in conjunction with Kinect SDK beta through System.Speech API, but results will be more error prone because only the Microsoft.Speech API has a language pack calibrated specifically for kinect characteristics. We know this current limitation is not ideal, but we hope to improve the state of the world over time with each release.

    Thanks for your understanding,
    Eddy


    I'm here to help
    • Marked as answer by MuGMaNia Wednesday, June 22, 2011 5:40 PM
    Wednesday, June 22, 2011 7:45 AM

All replies

  • I've been trying to do this as well, but have not figured it out yet. The only thing I could find was the GrammarBuilder.AppendDictation(), but the API doesn't really explain how to use it. I hope it's supported in the microsoft.speech api.
    Tuesday, June 21, 2011 1:30 PM
  • Unfortunately, Microsoft.Speech API does not support AppendDiction() because it doesn't support DictationGrammar.

    I have successfully implemented AppendDiction("spelling") using System.Speech API, but the recognition quality is way below par.

    I wish I could use either Windows Speech Recognition profiles, or at least be able to train the model.

    Tuesday, June 21, 2011 8:51 PM
  • MuGMaNia's impression is correct. It is possible to support dictation scenarios in conjunction with Kinect SDK beta through System.Speech API, but results will be more error prone because only the Microsoft.Speech API has a language pack calibrated specifically for kinect characteristics. We know this current limitation is not ideal, but we hope to improve the state of the world over time with each release.

    Thanks for your understanding,
    Eddy


    I'm here to help
    • Marked as answer by MuGMaNia Wednesday, June 22, 2011 5:40 PM
    Wednesday, June 22, 2011 7:45 AM