Menu

Sphinx4 vs Google SP system

bic-user
2012-01-11
2012-09-22
  • bic-user

    bic-user - 2012-01-11

    Greetings to all !!!
    First of all, lots of thanks to this project developers! I was extremely
    surprised to found fully-distributible code (vocabulary, processing
    algorithm). And there is a question. I'm searching for speech recognition
    system that can be used for the project with few key-words and rest are
    dictated. Whole phrase is gonna be something like 5-10 words. Android lib by
    google can't give enough accuracy (accent spoiling all intellegibility). And
    they do not give an opportunity to improve recognation accuracy. Your project
    looks very attractive in this way, but what accuracy I can get spending for
    example two month on adaptation and code researching? This question can look
    simple but it's very important for me to make right decision at the beginning
    of the project, so if you be so kind give all information: numbers, maybe
    demos I can take a look on..

    Thanks in advance.

     
  • Nickolay V. Shmyrev

    Hello

    To get more detailed information on CMUSphinx accuracy, usescased and the ways
    to improve the system you need to provide more information about the task
    itself:

    1. What kind of language do you want to recognize
    2. How audio will be deliverd to speech recognizer
    3. What meta-information for accuracy improvement can you provide
    4. Which accents do you have in mind

    It's sort of hard to advise you just on the information you provided.

    To let you know CMUSphinx is successfully used by many commercial vendors who
    solve real needs of their customers and solve them well. However, most of them
    put years of development on integration of the CMUSphinx core into speech
    recognition solutions.

     

Log in to post a comment.