Menu

Phoneme recognition without probability languange model

Help
2017-06-30
2017-07-03
  • Latio Notelly

    Latio Notelly - 2017-06-30

    Hello everybody!

    I need to get the phonemic composition of the word from the * .wav file. I will make some analytics on this set of phonemes (the task of preliminary determination of health problems by speech).

    Already now I can use CMU Sphinx with the language model to parse words into phonemes. And it works well. But the analyst gets blured because of the use of a probabilistic language model: this improves the quality of recognition by excluding from the search of incredible combinations of phonemes.

    Therefore, I need to get an accurate phonemic composition. I understand that this is a more resource-intensive procedure and the results may turn out to be less qualitative. It suits me, because I can drop the worst results.

    If I could get all the potential variants of the phonemic composition of the word in a file - it would be even better.

    I searched and read a lot, but I did not find a suitable method.

    How can I implement this with?

    Please help, thanks!

     
    • Nickolay V. Shmyrev

      The suitable method is to write any possible letter-phoneme mapping and then just go through all different combinations.

       

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.