Newsgroups: comp.speech
Path: pavo.csi.cam.ac.uk!doc.ic.ac.uk!agate!howland.reston.ans.net!sol.ctr.columbia.edu!sol.ctr.columbia.edu!mhall
From: mhall@occs.cs.oberlin.edu (Matthew Hall)
Subject: A little cepstral question.
Message-ID: <MHALL.93May4124551@occs.cs.oberlin.edu>
Sender: nobody@ctr.columbia.edu
Organization: Oberlin College Computer Science
Distribution: comp.speech
Date: Tue, 4 May 1993 17:45:51 GMT
X-Posted-From: occs.cs.oberlin.edu
NNTP-Posting-Host: sol.ctr.columbia.edu
Lines: 33

Hello,
	I have heard a lot about cepstral analysis in regards to
speech processing.  Right now I am just using the results of an FFT,
for the patterns I create, however I would like to increase accuracy a
little.  The cepstral is supposed to imitate the way the ear hears,
which should make recognition accuracy better.

Here are some questions:
The cepstral, as I understand it, is begotten by taking the FFT of a
signal, taking the complex logarithm of each frequency component, and
doing an IFFT on the result, i.e.
Cepstral = IFFT(Log(FFT(x))), where x is, say, a real valued sequence.
Now the FFT takes x from the time domain to the frequency domain, but
the IFFT takes it back to the time domain.  How do I interperet the
cepstral in a meaningful way if it is in the time domain?

Second, for processing concerns, may I disregard the phase info when
taking the cplx logarithm?  since log(z) = ln(|z|) + i*arg(Z), and
arg(z) is just the phase, can I simply use log(z) ->ln(|z|), and take
the IFFT of that, or is that a no-no?

I have looked at several DSP books, one which covers the cepstral, but
it isn't very clear about what to do with it (well, I'm not an
engineer, so I'm not sure about half of what its talking about)

Thank you for any help you can provide,
-matt
--
-------------------------------------------------------------------------------
Matt Hall.    mhall@occs.oberlin.edu
              (216)-775-6613 (That's a Cleveland Area code. Lucky Me)
  
"Life's like mayonnaise soda"	-Lou Reed
