Newsgroups: comp.speech
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!howland.reston.ans.net!pipex!uknet!comlab.ox.ac.uk!newshost!ramin
From: ramin@psy.ox.ac.uk (Ramin Nakisa)
Subject: Re: information about mel process
Message-ID: <RAMIN.95May2191118@phoenix.psych.ox.ac.uk>
In-reply-to: sintes@geant.cica.fr's message of 20 Apr 1995 08:27:36 GMT
Organization: O.U. Dept. of Experimental Psychology
References: <3n55to$qtm@pelvoux.cica.fr>
Date: 02 May 1995 18:11:18 GMT
Lines: 31

In article <3n55to$qtm@pelvoux.cica.fr> sintes@geant.cica.fr (The Greatest Poet of the World) writes:

>   hi everybody
>   I must use the Mel scale to improve the recognition.
>   I would have some information about Mel process ,Mel cepstral
>   coefficient,Mel-scale frequency axis

You may like to take a look at a beautiful article by Tony Robinson

 http://squid.eng.cam.ac.uk:80/Reports/ajr/icslp90

He has tried many different preprocessors and determined their affect
on phoneme recognition accuracy using an SRN on the TIMIT database. In
the abstract he says:

"The paper concludes that the features do not make a significant
contribution and that the spectral domain representations, independent
of their derivation, are better suited to this task. However, we find
that the recogniser was relatively insensitive to preprocessor and
changes in the architecture and training of the recogniser are more
significant."


--
+--------------------------------------------------------------------------+
| Ramin Charles Nakisa,                    Tel:   (0)1865 271400           |
| Department of Experimental Psychology,   FAX:   (0)1865 310447           |
| University of Oxford,                    EMAIL:                          |
| South Parks Road,                               ramin@psy.ox.ac.uk       |
| Oxford OX1 3UD.                                 mbrcn@seqnet.dl.ac.uk    |
+--------------------------------------------------------------------------+
