Menu

Fixed point frame power calculation

Help
Pankaj
2010-09-24
2012-09-22
  • Pankaj

    Pankaj - 2010-09-24

    I observe that there is no fixed point implementation for frame power
    calculation and noise level calculation in pocketsphinx in the file
    cont_ad_base.c

    The functions cont_ad_frame_pow() and find_thresh(cont_ad_t * r) have a couple
    of floating point calculations: ---
    i = (int32) ((10.0 * (log10(sumsq) - log10((double) spf))) + 0.5);
    and r->noise_level = (int32) (r->noise_level + r->adapt_rate * (th -
    r->noise_level) + 0.5);

    Was there any particular reason for not implementing fixed point version of
    these equations.
    Can somebody suggest some literature for implementing these in fixed point.

    Regards
    Pankaj

     
  • Nickolay V. Shmyrev

    Was there any particular reason for not implementing fixed point version of
    these equations.

    They are not critical for application performance.

    Can somebody suggest some literature for implementing these in fixed point.

    Something like that should work. You then need to multiply it on constant

    http://embeddedgurus.com/stack-overflow/2008/05/integer-log-
    functions/

     

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.