Re: Speech rate modeling in synthetic speech

[Thomas G Brennan <g_brennantg@xxxxxxxxxxxxxxx>]

Any synthesizer which allows you access to its phoneme generation system can be
programmed to give pretty normal rate to speech but this means you have to
program every word and sound.  Text recognition, lpc, and similar schemes either
stretch the lengths of vowels for slower speech or chop sounds (interrupt) to
create faster speech.  If interruptability is turned off on the synth unless you
individually program sounds you will be limited by the speed of the system's
internal rules and dictionaries.  Even if you program for the synthesizer this
will be problematic but can be overcome to a large extent.  You will almost find
programmer's toolkits a must for programming as specific as you want.  I have
designed and written programs for a few sunthesizers that help shortcut that
process but programming everything from scratch is very time consuming and takes
a good deal of work.  Depending upon the internal dictionary bein used by the
synth or dictionaries being added at program load-up you may or may not want to
disable the on board system rules.

Tom


Tom Brennan  KD5VIJ, CCC-A/SLP
web page http://titan.sfasu.edu/~g_brennantg/sonicpage.html