Newsgroups: comp.speech
Path: lyra.csx.cam.ac.uk!doc.ic.ac.uk!daresbury!keele!uknet!dcs.gla.ac.uk!unix.brighton.ac.uk!jwr
From: jwr@unix.brighton.ac.uk (reed)
Subject: Speech synthesis and phonemes - help!
Message-ID: <1994Apr29.110924.14879@unix.brighton.ac.uk>
Organization: University of Brighton, UK
X-Newsreader: Tin 1.1 PL4
Date: Fri, 29 Apr 1994 11:09:24 GMT
Lines: 44

I am currently undertaking my final year project at the University of Brighton
. The project involves development of software to control a custom built speech
board (based on the RS ic263 chip). The chip is now quite old but contains
a set of 64 phonemes/allophones. Various parameters for each phoneme can be
set such as duration, pitch, amplitude, rate of articulation, etc. The
data sheet claims natural sounding speech can be produced.

The main objective of the software is to take text as input, convert this to
phonemes and send these to the chip with values for the above parameters.
Firstly, I have some rules to transcribe text to phonemes but these are limited
 and therefore I have too many exception words (there is a dictionary kept
 to store these effectively ), can any one supply any rules for the
 transcription of text to phonemes or point me to where I might locate some.
 
 Secondly, I wish to add some intonation and make the speech better than
 simply monotonic, therefore I need to intelligably assign values to
 the phoneme parameters to do this. I also wish to take into account
 punctuation in text. Can anyone suggest the most important paremeters
 to manipulate that will produce the most noticable affects, I
 believe some parameters are attributes of the phonemes themselves and 
 others affect the transition between phonemes. Additionally is anyone
 able to supply any algorithms to calculate appropriate values for the
 parameters. The parameters that can be manipulated are phoneme duration,
 phoneme amplitude, filter frequency (base frequency), rate of speech,
 rate of articulation, pitch. Is it possible to break a word down into
 its syllables using some algorithm rather than a dictionary? if so
 does anyone have such an algorithm alvailable? Also in the English 
 language there are a limited set of allowable consonant letter groups
 that can be used to identify real words from acronyms, can anyone 
 supply them?
 
 Lastly, at present the overall rate of speech is dependent on the speed 
 of the PC (the faster the PC the faster the data is sent to the board
 ) any suggestions to make the speech rate processor speed independent.
 
 Development of the software is in Visual Basic (front end) and C (
 residing in DLLs). If anyone can assist me with any of the above I would
 be very grateful if they could mail me with there suggestions or post a 
 reply here, 
 
 thanks a much,
 
 Jonathan.
 
