Date:     Oct. 29, 3:30 pm, 3024 Engineering Hall

Title:    In the Future, All Computers Will Sound
          Like They Come from Wisconsin

Speaker:  Paul Milenkovic, ECE Department

Abstract:

The old Star Trek had the computer speaking in a stilted monotone.
Kubrick foresaw advances in speech science that had his HAL 9000
speaking in a natural voice.  In real life Stephen Hawking, disbabled
with neurological disease, speaks with a machine voice somewhere in between.
His DECTalk synthesizer is the life work of Dennis Klatt, a Wisconsin
native who embodied his Scandinavian-influenced lilt into the the
synthesis rules.

Further advances in naturalness may require knowing the underlying
vocal tract shapes.  The University of Wisconsin X-ray Microbeam Speech
Production Database contains 20 hours of speech cumulated from 57 talkers,
primarily UW undergraduate students.  This will ensure that future
computers will continue to have a Wisconsin accent.

My work concerns determining vocal tract shape from the combination
of the x-ray measurements taken with the x-ray system and acoustic
features.  I will discuss the theoretical basis for determining vocal
tract shape from acoustic information, why acoustic data alone does
not specify a unique shape, and results of matching both the x-ray and
acoustic data.  I will demonstrate computer animations of simultaneous
sound and speech movement, and I will demonstrate the operation of
a computer model of vocal articulation.


==============================================
***  NOTE LOCATION: 3024 Engineering Hall  ***
==============================================