Newsgroups: comp.speech
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!fas-news.harvard.edu!newspump.wustl.edu!news.ecn.bgu.edu!news.moneng.mei.com!uwm.edu!spool.mu.edu!howland.reston.ans.net!pipex!uknet!cix.compulink.co.uk!usenet
From: jhaseler@cix.compulink.co.uk ("John Haseler")
Subject: Re: speech enhancement
Message-ID: <D6zuLH.L96@cix.compulink.co.uk>
Organization: Compulink Information eXchange
References: <3m6hc6$d1g@eiger.ceet.niu.edu>
Date: Thu, 13 Apr 1995 22:03:16 GMT
X-News-Software: Ameol
Lines: 13

In my opinion, you can use all sorts of good measurements but it is very 
hard to persuade users that 'better' is better (and presumably that is 
what matters in the end).  The only thing I really trust is a mass of 
audibility trials, like Diagnostic Rhyme test for tens of listeners, 
which is hardly practicable.  Mechanical things measure what machines 
like to look at, if you can persuade yourself this is relevant.  Using a 
little data many times, and listening, persuades you it is getting better 
- when you are just learning.

I wish I knew of a good answer, but haven't yet met one I really believe 
in.

John Haseler
