Newsgroups: comp.dsp,comp.speech
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!uhog.mit.edu!bloom-beacon.mit.edu!gatech!psuvax1!news.pop.psu.edu!news.cac.psu.edu!newsserver.jvnc.net!newsserver.egr.uri.edu!ramli
From: ramli@ele.uri.edu (Ramli)
Subject: Tutorial on Wavelets; articles on speech recognition
Keywords: wavelets, tutorial, speech recognition
Sender: news@egr.uri.edu (USENET News System)
Organization: URI Department of Electrical Engineering
Message-ID: <D0x1py.35y@egr.uri.edu>
Date: Fri, 16 Dec 1994 18:22:45 GMT
Lines: 63
Xref: glinda.oz.cs.cmu.edu comp.dsp:15311 comp.speech:4087

There are quite a few people who think that wavelets are
the greatest invention since sliced bread.  Whether or
not you subscribe to this view, you might still be interested
in taking a look at the following intro' article:

  Wavelet Transforms: An Introduction
  P.M. Bentley and J.T.E. McDonnell
  
  Electronics and Communication Engineering (IEE)
  Vol. 6, No. 4, August 1994, pp. 175--186.

There is also a new book on the topic: "A Friendly Guide
to Wavelets" by G. Kaiser of U. Mass. at Lowell.  The
publisher is Birkhauser, Boston; 300 pp., 1994,
ISBN 0-8176-3711-7.  (No, I don't know the author, nor
is Birkhauser giving me any kickback for this publicity;
just a friendly service as part of the 100 hrs. community
service that I have been sentenced to :-))  The spiel in
the catalog that made its way to my friend's mailbox reads:

  This volume consists of two parts.  The first part,
  forming Chapters 1--8, is designed as a textbook for
  an intro one-semester course on wavelet analysis and
  time-frequency analysis aimed at graduate students
  or advanced undergraduates in science and engineering.
  Each of the first eight chapters ends with a set of
  straightforward exercises designed to drive home the
  concepts just covered, and the graphics should further
  facilitate absorption.  The second part, forming
  Chapters 9--11, represents original research and is
  written in a more advanced style.  This section can
  be used as a textbook for a second-semester course
  or, when combined with chapters 1 & 3, as a reference
  for an advanced research-level seminar.

I haven't seen the book myself and so have no comments
on the book.  But it may be well worth taking a look at.

There are a couple of interesting articles on current
products in speech recognition in Forbes ASAP (the
technology supplement to the Forbes magazine) dated
Dec. 5 1994.  The first article is entitled "News Talk
at the New York Times" (pp. 86--88) and features the IBM
Tangora system.  The second one is "The Lady and the Dragon"
(pp. 89--90), featuring the DragonDictate.  The following
paragraph from the second article sums up accurately the
challenges facing present-day recognizers:

  Taylor says communicating with the computer requires
  precise enunciation and verbal control.  This is why
  training of new users is so critical, say experts.  "If
  I talk too fast, it doesn't work.  If I talk too loud,
  it doesn't work.  If I am in a bad mood, a happy mood,
  stressed out---anything that changes the tone of my
  voice---the Dragon makes mistakes," say Taylor.

This article also lists four or five similar products,
with names and addresses of where to get them.


ramli.


