Menu

Using Keyword spotting in Pocketsphinx

Help
Leo
2016-07-04
2016-07-06
  • Leo

    Leo - 2016-07-04

    Hello everyone,

    As I would like to use keyword spotting provided by Pocketpshinx for my project, I was wondering if I could use it for 108 different french phonemes (e.g. la, br, ma, st, sk, le, o etc), and possibliy avoiding false alrams. Could any one of you please let me know whether this is possible? What would be the optimal way to tune each of the phonemes, such that having uniqe thresholds for each one? Would be the optimal way if I try to look for the optimal threshold by testing each phoneme individually until the right threshold is found?

    Thank you very much in advance for any suggestion.

    Leutrim

     
    • Nickolay V. Shmyrev

      It is not possible.

       
      • Leo

        Leo - 2016-07-04

        Hello Nickolay,

        Thank you very much for your fast reply.

        As you know, I had problems to do this using phoneme recognition, then I thought I might use this functionality of keyword spotting.

        Can you please give me any suggestion how can I solve this problem (as you know I am trying for a long time to solve this)? When I used phoneme recognition, I had additional phonemes appearing (I know it was normal to have them there, but I still do not know why?)...

        Thanks a lot!

        Leutrim

         

        Last edit: Leo 2016-07-04
        • Nickolay V. Shmyrev

          This might be an issue with French model. You can try to reproduce this issue iwth English model first.

           

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.