Menu

Pocketsphinx recording and VAD

Help
Duki
2011-08-19
2012-09-22
  • Duki

    Duki - 2011-08-19

    Hi, I am trying to implement pocketsphinx like VAD (voice activity detection)
    in my project and have been trying to dig up in the code how pocketsphinx
    recordes data.

    I have seen that during VAD it takes recorded data from predefined buffer that
    are supposedly filled by recording data but i can't figure out where are they
    being filled? And I am wondering as the program is doing lots of callculation
    oll the time how does it record audio at the same time?

    I would be in need of that too and i hope it is done by software somehow
    becouse my project involves VAD and MFCC extraction an a microcontroller and i
    hope to get functionality like sphinx, i need speed and eficiency to hopefully
    get a real time implementation.

    Hope some of you can help me. If you could exoplain how it recordes and does
    VAD at the same time, oh and how does it transer data exactly between the two.

    Thank you in advance.

     
  • Nickolay V. Shmyrev

    how it recordes and does VAD at the same time, oh and how does it transer
    data exactly between the two.

    Pocketsphinx works in multitasking operating system. While kernel is recording
    sound to a buffer pocketsphinx works on detecting voice in it. It requests for
    a new buffer periodically and OS kernel fills it through the alsa library for
    example.

    Usually recording is implemented by means of hardware interrupts. You only get
    a buffer of audio when it's full periodically. Between the interrupts you can
    process the buffer data, in particular you can detect voice in it. In
    multitasking operating system this functionality is hidden behind the OS
    interface.

     

Log in to post a comment.