Using speech recognition effectively for real-time captioning? RRS feed

  • Question

  • Hello,

    I feel like I am asking a silly question here, but I'll go ahead anyway. I understand that the Kinect currently understand a list of commands. But I if were to make a database of common grammer pharses, such as


    Which would look like this in the grammers builder:

                        var grammars = new Choices();

    Would it be possible for the Kinect speech recognition engine to effectively recognise all the above spoken words in one single speech without need to take a long pause between each spoken word? I understand I could simply just list "Hello how are you" as a single command, but I am wanting to do a mix of spoken words to produce near enough real time captioning.

    - Dan



    Tuesday, October 18, 2011 10:06 PM


  • Your scenario is possible, but I expect that you would have a larger error rate if you don't pause much between words. The speech recognizer works best for commands and less well for dictation scenarios.


    I'm here to help
    Friday, October 21, 2011 2:07 AM