Newsgroups: sci.electronics,comp.speech
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!howland.reston.ans.net!gatech!bloom-beacon.mit.edu!news.tamu.edu!news.utdallas.edu!corpgate!bcarh189.bnr.ca!nott!cunews!freenet.carleton.ca!FreeNet.Carleton.CA!al904
From: al904@FreeNet.Carleton.CA (Rick Sloan)
Subject: ==== text to speech help ====
Message-ID: <CzAu6M.8u4@freenet.carleton.ca>
Sender: al904@freenet.carleton.ca (Rick Sloan)
Reply-To: al904@FreeNet.Carleton.CA (Rick Sloan)
Organization: The National Capital FreeNet
Date: Tue, 15 Nov 1994 07:59:10 GMT
Lines: 72
Xref: glinda.oz.cs.cmu.edu sci.electronics:108127 comp.speech:3801


Just found this interesting piece of email on gopher. Can anyone elaborate
or know of the BYTE article they talk of?????

--------------
This section is from the document '/SPIB/news/comp.speech/1993.12'.

>From rice!newsfeed.rice.edu!cs.utexas.edu!uwm.edu!spool.mu.edu!cass.ma02.bull.com!steamer.clam.com!spdcc!jti.com!ci-pioneer!icarus!mirons Tue Dec 21 05:52:38 1993 
Path: rice!newsfeed.rice.edu!cs.utexas.edu!uwm.edu!spool.mu.edu!cass.ma02.bull.com!steamer.clam.com!spdcc!jti.com!ci-pioneer!icarus!mirons
From: mirons@icarus.ci.net (Michael A. Irons)
Subject: Re: Text to Speech on a PC
References: <lawrence.756370744@s1.elec.uq.oz.au>
Organization: NewEngland Community Internet; Boston, Ma. USA - Private Node
Date: Tue, 21 Dec 1993 05:52:38 GMT
Message-ID: <CIDEzr.3qy@icarus.ci.net>
Lines: 42


	For peoples info. There is a diphone based T-to-S program.
written in C, you should be able to get source. If you can _Please_
make sure that the code, and diphone library gets put out for
anonymous FTP and the location gets into the FAQ. If you end up
talking to Rick, the lab manager at the time say that _I_ am asking
this as the author.  (wow, I've been on your end of that statment all
this time.  Feels weird writing 'as the author' of a software package
for the first time on the net)

There are 13 R.E.C.s (Rehabilitation Engineering Centers) funded by
the US government. Each has a seperate goal. I worked at the one
_formerly_ at Tufts NewEngland Medical Center here in Boston
(1986-1988). The grant moved because Tufts got to gready, wanted to
much of it for themselves. This RECs focus was augmentative
communitation (builder of TIC - Tufts Interactive Commmunicator). One
of the projects was a proof study on the feasibility of a diphone
based T-to-S that would be cheeper then DEC-talk. When they moved they
had just made a female diphone library for a comercial client.

I wrote the driver program and the librarian. Someone else compiled a
rule set and extracted diphones from a set of stimulus words. The
diphones were converted to Linear Predictive Code and stored in the
library.  Ditalk took typed text, ran it through the rule set to
convert the text into a stream of diphone symbols. They were copied
from the library, concatenated, and sent to a TI board for realtime
LPC to sound. The whole thing was based on a idea from a Byte artical :)

Though the conditions of the recording were not the best (ex - no
noise reduction and done in the lab) you could recognize the person
who did the stimulus words. If you can track down this, don't expect
earth shatering quality. 

As the government (read your taxes) paid for this it should be
obtainable.  Locating is is left as an exercise for the reader. I'm
going to be homeless in two weeks so don't bother trying to email me
and expect a reply :( x 1E(1E100)
-- 

				Mike Irons

			mirons@Icarus.CI.NET

--------------

ANY IDEAS WHERE TO GET INFO????? OR ANY OTHER LEADS.......

**RICK**


--
* Rick Sloan (613)723-1427 * P-57 Mustang & P-51 Mustang Virtual C Processors
* IntelliHome Controls     *  Intelligent Home Controls for disabled persons
* 15 - 8 Deerfield Drive,  *    Embedded systems design, 87C750, PIC16C54
* Nepean, ON, K2G 3R6      *     Dallas Touch Memory, SmartCard designs
