Newsgroups: comp.speech
Path: pavo.csi.cam.ac.uk!warwick!pipex!uunet!munnari.oz.au!bruce.cs.monash.edu.au!trlluna!titan!titan!duke
From: duke@titan.trl.OZ.AU (Peter Duke)
Subject: Re: Dirty Sox
Message-ID: <1993Nov16.024625.276@trl.oz.au>
Sender: root@trl.oz.au (System PRIVILEGED Account)
Organization: Telecom Research Labs, Melbourne, Australia
References: <CGB4A6.CIn@watserv1.uwaterloo.ca>
Date: Tue, 16 Nov 1993 02:46:25 GMT
Lines: 25

anderson@crypto2.uwaterloo.ca (Bill Anderson) writes:
> I have been using sox to convert 16-bit 16kHz TIMIT
> speech files to 8-bit 8kHz ulaw PCM.  The problem is
> that sox seems to be doing a very dirty job of either
> the filtering or the rate conversion.  The PMF
> (probability mass function) for a fair sample of data
> reveals significant discontinuities at intervals of
> 16 output levels.  This is very much unexpected and 
> very likely in error since the PMF of the original 
> TIMIT data is quite smooth.
> 
> Can anyone tell me how to fix this problem or point
> me in the direction of a better conversion routine?
> 
> Many thanks,
> 
> Bill Anderson
>
Are you sure that its not due to reducing the 16 bit linear PCM to the
piece-wise linear log PCM implied by u-law?

Peter Duke
Telecom Research Laboratories
p.duke@trl.oz.au
 
