[Speex-dev] extracting voice from sound data

[Oleg Mamaev]

Hello All!

I have very important question (sorry for my bad English):

How can I get sound data (in speex format) which have amplitude larger than
specified value?
I.e. I need getting speech which volume is larger than specified value.
For 16-bit linear PCM-encoded data this is very simple. Because each sample
contains 16 bit sound data.
And I compare this 16 bit value with specified value for each sample. For
example, I assume that sample would part of voice,
if their sample size value > 10000 or < -10000 (maximum value of sample size
= 32767 (2x15-1) and minimum value = -32768 (-2x15)).
But how I doing similar operation with data in speex format rather than in
PCM? Probably, is there some formula for this operation?
My data encoded with jspeex (CBR, 8KHz (narrowband mode), mono, little
endian, quality = 7).

Thanks In Advance.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.xiph.org/pipermail/speex-dev/attachments/20081018/94d4bba3/attachment.htm