Menu

sphinx3 regression test

Help
Arslan
2007-05-31
2012-09-22
  • Arslan

    Arslan - 2007-05-31

    Hi guys,

    I'm trying to perform speech recognition using sphinx3.
    I have downloaded the sphinx3-0.6.3 release and compiled it as you explain
    in your documentation.
    I've seen that you had regression test, so I tried to check wether the
    program was working well or not.
    For every regression test I do this:
    ./test-decode-simple.sh
    And I always get the result:
    test FAILED.
    Obviously the recognition doesn't work. Do you have any idea what I might have done wrong.
    I'm totally lost, I don't know what to do.
    Your help would be very appriciated.

    Thank you
    Best regards.

     
    • Nickolay V. Shmyrev

      Without logs it's impossible to help you. This script should create a file test-decode-simple.out in the same directory, share it's somewhere.

       
    • Arslan

      Arslan - 2007-05-31

      Hi,
      Thank's for answering.
      I actually have solved my problem.

      here was the problem.
      I was launching the tests from ~/src/tests/regression
      Here is the log I got:
      src/tests/regression/../../.././libtool: src/tests/regression/../../.././libtool: No such file or directory

      It actually need to be launch from:
      ~/Sphinx3/sphinx3-0.6
      and do
      src/tests/regression/test-decode-2ndstage.sh
      otherwise the program can't find what it is looking for.

      Thank's for answering.
      I have a new question though ;)
      I'm willing to do a free speech recognition program.
      I was doing this with sphinx4 using rm1 it was working pretty well.

      I would like to do the same thing with sphinx3 but it is kinf of hard to configure, with all these arguments...
      Do you know if I can find somewhere where there are good example on how to configure it correctly, to optimize the recognition.

      Thank's again.
      Best regards.

       
      • Nickolay V. Shmyrev

        It depends on what are you going to recognize and how.

         
    • Arslan

      Arslan - 2007-05-31

      Hi again,

      I am going to do a free speech recognition.
      The user can say wathever he wants.
      The acquisition will be perform with a microphone.
      And then the audio file will be converted using sox.

      I had noticed that the rm1 acoustic model gave better recognition results than hub4 when I was using sphinx4. This is why I'm planning on using rm1. I am aware that the system won't be able to recognize every words, but at least it will do less mistake on simple words.

      I think I need a trigram langage model.

      Thank's for answering again.

       
    • Arslan

      Arslan - 2007-06-01

      I forgot to precize that the conversion of the audio file will be done from .wav
      to .raw.

       

Log in to post a comment.

Want the latest updates on software, tech news, and AI?
Get latest updates about software, tech news, and AI from SourceForge directly in your inbox once a month.