3 Replies Latest reply: Feb 24, 2010 11:13 AM by 843802 RSS

    Implementing Speech Recognition for simple String and Integer Analysis

    843802
      Im working on a Science Fair project and I would like to implement a Speech recognition system so that the users can interface with the test program by speaking. So far, I have tried to use Sphinx but the implementation is giving me problems. Is there a better way to accomplish what I'm doing, where can I find it, and how can I implement it? All I need is to accept a direction, then an integer, and then repeat because it will be in real time.
        • 1. Re: Implementing Speech Recognition for simple String and Integer Analysis
          843802
          I had problems with Sphinx in the past too.
          I tried extending a simple demo program to recognise phone numbers and to read them back using FreeTTS speech synthesis.

          But the results were hit and miss. Sometimes it would work nicely, then I'd try looping around to do it again, and it won't do recognise the number the second time. It was like talking to a partially deaf person... you had to get the timing right.

          Sorry, no solutions for you, just confirming "issues".
          I had tried using a headphone with built-in mic, just like a call-centre headset, in an attempt to improve / reduce background noise. But wasn't much better.

          I imagine in a science fair you're going to have a lot of ambient noise, making it even worse. Maybe some more robust input device is needed for direction... a joystick springs to mind.

          regards,
          Owen
          • 2. Re: Implementing Speech Recognition for simple String and Integer Analysis
            843802
            My project tests the efficiency of different interfaces. I was going to add joystick as well as the standard mouse and keyboard. But what would make it an excellent project is if I could have voice as an interface.
            • 3. Re: Implementing Speech Recognition for simple String and Integer Analysis
              843802
              hi, for simpler voice command entry, you may try vText.jar. It has recording (5s), filter (wavelet), comparison functions

              Free version is available at http://www.basic-signalprocessing.com/downloadVText.htm, developed in 2009

              u may need write ur own loop functions to effect repeat. the class should be instantiated before use, best time is when application is started.