Newsgroups: comp.speech
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!noc.near.net!analog.com!analog.com!nwd2sun2.analog.com!az
From: az@saturn.analog.com (Alex Zatsman)
Subject: Re: LD-CELP (G.728) speech coder
In-Reply-To: mo@kom.auc.dk's message of 24 Oct 1994 10:51:55 GMT
Message-ID: <AZ.94Oct28203704@saturn.analog.com>
Sender: usenet@analog.com
Organization: Analog Devices Inc, Norwood MA, USA
References: <MO.94Oct24115155@orion.kom.auc.dk>
Date: Sat, 29 Oct 1994 01:37:04 GMT
Lines: 35

In article <MO.94Oct24115155@orion.kom.auc.dk> mo@kom.auc.dk (Morten Olesen) writes:

>    Has anyone tested the Low Delay CELP coder mentioned in the FAQ:
> 	   ftp://dspsun.eas.asu.edu/pub/speech/ldcelp.tgz
>    with the official CCITT test sequences?
> 
>    My preliminary tests on speech data yield a speech  quality worse than the GSM coder,
>    which is unexpected.
> 
>    Please  inform me either in  this newsgroup or directly by  e-mail if you have
>    any patches to the mentioned software or any other supplementary information.
> 

While I'm not 100% sure, it does sound  like you are talking about the
implementation I put in the public  domain some  time ago --  at least
the file name looks familiar.

If  that is the  case, I would like  to reiterate the  points  already
stated in the accompanying README file:

My   implementation  has NOT  been  tested with  the official vectors.
Indeed, the  whole  program  was   written mostly as   a test  for our
floating point  C  compiler. Please do   not judge  the   G.728 by the
quality of my  implementation since the latter  is quite likely  to be
buggy.

Also, please note "no warranty" disclaimers in the source and the
README file.

And lastly: if you find any bugs and fix them -- I'll appreciate if
you share them with me and others. Thank you.

--
Alex Zatsman,     Analog Devices, Inc., (617) 461-3729
alex.zatsman@analog.com
