[Lab] Speech rec for embedded device

[Jean-Marc LeBlanc]

Does any one know how speech recognition works? what I understand so far,

sample 0.96 sec or audio
filter it
fft to generate a vecotor
then they use the hidden Markov model to determine the best match.

I am having trouble understanding hmm and how it works with speech
rec.  I am hopping some one might have done some thing like this in
the past.
It seems to be similar to pattern detection. If i am not mistaken
there are a few CV expert in the list/.

Really I am looking for a good read that is not limited to some
library or a frame work

any info or tips would help.  I did do my own research on Google
every thing I found was tutorials on how to use specific framework or
tool kit.  I don't really mind tool kits but I would like to apply the
bsd or mit license to my code. though even if its a restrictive
license it might be a good learning resource.

I think my best bet would be to port this project
http://www.isip.piconepress.com/projects/speech/html/license.html.

thanks for helping me brain storm


Jean-Marc Le Blanc
---

"Do you pine for the nice days of Minix-1.1, when men were men and
wrote their own device drivers?" Linus Torvalds