Menu

MLLR in Sphinx 4

Help
2008-11-27
2012-09-22
  • Gabriel Skantze

    Gabriel Skantze - 2008-11-27

    Is it possible to use adapted acoustic models (MLLR) in Sphinx 4? I haven't found any description on how to do this.

     
    • Gabriel Skantze

      Gabriel Skantze - 2008-11-27

      Yes, I would be very interested. Would it be much work?

       
      • Nickolay V. Shmyrev

        Well, you probably want to start with something simplier. The little code for LDA handling is located in sphinxbase/src/libsphinxbase/feat/lda.c. That code should be ported in Java and embedded into the frontend. Such transform will give you a significant accuracy improvment and will speedup the training of the model.

         
        • Nickolay V. Shmyrev

          Don't care about LDA, I've just committed the code to load it. It still has bugs though and doesn't work as expected.

           
    • Nickolay V. Shmyrev

      No, this feature is not supported yet. We can help you with the implementation if you are interested. Also it would be nice to support LDA/MLLT as it was done in pocketsphinx.

       

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.