Newsgroups: comp.robotics
Path: brunix!news.Brown.EDU!qt.cs.utexas.edu!yale.edu!newsserver.jvnc.net!darwin.sura.net!sgiblab!munnari.oz.au!metro!seagoon.newcastle.edu.au!wombat.newcastle.edu.au!asrgf
From: asrgf@wombat.newcastle.edu.au
Subject: Re: Text to Speech
Message-ID: <1993Feb17.223758.1@wombat.newcastle.edu.au>
Lines: 53
Sender: news@seagoon.newcastle.edu.au
Organization: University of Newcastle, AUSTRALIA
References: <1993Feb13.042426.10028@magnus.acs.ohio-state.edu> <C2EuGz.2E3@cs.uiuc.edu> <1993Feb14.012902.20900@ncsu.edu>
Date: Wed, 17 Feb 1993 11:37:58 GMT

In article <1993Feb14.012902.20900@ncsu.edu>, djbarnes@eos.ncsu.edu (DONALD JAMES BARNES) writes:
> 
> 
> 
> Someone a while back wanted some info on a chip that could turn ascii to 
> phonemes or ascii to speech or something like that.  I found a book today
> (don't remember the name) that had a circuit to turn text to speech.  I didn't
> get the part numbers, but it used a chip from General Instruments to do just
> that.  I didn't have enough cash on me at the time to buy the book, but plan
> to go back and get it next week.  Anyone wanting info should mail me and I'll
> respond as soon as I get the book.  BTW, it had a lot of other neat robotics
> circuits in it as well.
> 
> Donnie
> 
> ******************************************************************************
>                          |
> Donald Barnes	         |     	"You can think about it, but don't
> djbarnes@eos.ncsu.edu    |
> N.C. State University    |			do it."--Smokey
>                          |
> ******************************************************************************


 The chips you want are (were?) made by GI. One produced "phonemes", really
alloph, from hexadecimal input codes. Each of the approx. 64 allophones
had it's own distinct code. You could put together reasonably clear speech
by sending the chip a string of these codes corresponding to the allophones
of the utterance required. I cannot rember the name of the chip but I have 
one at work and I can find out tomorrow.

Text to speech was performed by another chip, the GI cts256a-al2. This could be
combined with the previous device in a circuit which took as input an ASCII
text string version of the utterance. i.e. a word, sentence or phrase. 
The cts256a-al2 was a microprocessor with an internal ROM containing the 
327(??) odd rules of the Naval Algorithm. It's output was the equivalent
allophone "string" for the synthesiser chip.

I did my Masters thesis on this topic during the early eighties. I built a
speech synthesiser using the Votrax Sc-01 phoneme chip, and a 6502 
microprocessor implementation of the Naval Algorithm. The latter was published
by the U.S. Naval Research Laboratory in the mid 70's for the Votrax phoneme
set. I also wrote a Z80 version. Resulting speech was slightly inferior to that
of the GI chip set, which was released several years later.

Does anyone know if Votrax chips are still available?

Ray Fairall
Department of Applied Science,
University of Newcastle.
N.S.W. Australia.
  

