Menu

pocketsphinx.. out of vocabulary word rec

Help
cornelyus
2010-11-22
2018-10-25
  • cornelyus

    cornelyus - 2010-11-22

    Hi..

    So, i was able to adapt an acoustic model for my voice and created a grammar
    with the sentences I needed to recognize.

    With pocketsphinx_batch, i could get almost 100% accuracy... seemed really
    good..

    Next step, started using pocketsphinx_continuous to try say a few sentences
    and see how it behaves..

    Everything I said ( that I EXPECTED to be recognized ) succeeded in being
    recognized.. the problem was when it started recognizing sentences, when i
    said random things, when people around me spoke, when they made some noise..
    when i cough...

    Did some searches here, and it seems this is a problem with out of vocabulary
    recognition...

    I am a bit stuck here now.. how could I go from here?

    I have my AM ready, a grammar.. i have to have a way of rejecting the sounds
    that are completely random...

    Maybe a way is recording again wav files for training AM with noise on the
    background?

    Thank you

     
  • Enas Ashraf

    Enas Ashraf - 2010-12-09

    can you tell me how did u build your grammar coz i'm facing the same problem
    here

     
  • cornelyus

    cornelyus - 2010-12-09

    Hi..
    For pocketsphinx i couldn't do it..
    For sphinx4 I could come up with a solution to the out of vocabulary
    recognition issue ...

    What were you interested in?

     
  • Enas Ashraf

    Enas Ashraf - 2010-12-09

    well, the case is that i want to build my own model to be used with
    pocketsphinx. the hmm and dictionary file as well.i can't use the pre-built
    model coz they r too large for mobile applications, so i need to to build a
    few words model and dictionary. that's the case.

     
  • Ayhan KÜÇÜKMANİSA

    i have same problem Did you solve this problem cornelyus?

     
  • cornelyus

    cornelyus - 2011-01-13

    well, the case is that i want to build my own model to be used with
    pocketsphinx. the hmm and dictionary file as well.i can't use the pre-built
    model coz they r too large for mobile applications, so i need to to build a
    few words model and dictionary. that's the case.

    Model.. are you talking about acoustic model? If so.. the tutorial ( http://c
    musphinx.sourceforge.net/wiki/tutorialadapt
    ) helped me adapting the example own acoustic model to fit
    my voice, and make it suitable for mobile application .

    As for the dictionary, you can also build your own.. http://cmusphinx.sourcef
    orge.net/wiki/tutoriallm?s
    =dictionary

    i have same problem Did you solve this problem cornelyus?

    you refering to "out of vocabulary" on pocketsphinx? If that's the case then
    no..

     

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.