Menu

When to train an Acoustic model?

Help
Sailesh
2008-10-31
2012-09-22
  • Sailesh

    Sailesh - 2008-10-31

    I'm new to Sphinx4 as well as speech recognition,i have installed sphinx4 and able to execute HelloWorld program. I couldn't understand when i have to train an acoustic model.If i change the jsjf grammar file it is still able to recognize words so i think the model is independent of grammar and the HMM model only considers phenoms that makes up the word.Am i right?
    If i plan to add more words to my dictionary or when my user base changes (e.g: my model is trained in US accent but my users have a different accent)then only i have to train an acoustic model?

     
    • Nickolay V. Shmyrev

      > If i change the jsjf grammar file it is still able to recognize words so i think the model is independent of grammar and the HMM model only considers phenoms that makes up the word.Am i right?

      You are correct.

      > If i plan to add more words to my dictionary

      As far as you use the same phones, you don't need new acoustic model

      > when my user base changes (e.g: my model is trained in US accent but my users have a different accent)t

      Depending on the difference in the accent you could either adapt existing acoustic model or train a new one.

       

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.