Adding an Extra Constraint To Help Recover Articulator Positions
Michelle Heard, Department of math and Computer Science, Grambling, LA 71245

Abstract:  Speech is the number one means of communication amongst humans. 
Due to today's technology, it would be a miraculous idea if one could be 
identified by what one is saying through the analysis of a computer.  The 
project's goal is to be able to recover the position of the lips, jaw, 
tongue, and other articulators from speech sounds alone.  To develop this
procedure several algorithms were used; such as, vector quantizing speech 
signals, then using multi-dimensional scaling to represent quantized code in a 
topographical map. The basic work was developed by the group of Dr. John Hogan,
Los Alamos National Laboratory.  An extra constraint is being developed by 
implementing a program to smooth the estimated positions of a topographical 
map and to produce a code book of those positions.  Multiple regression is
used to determine the correlation between the estimated and the recovered 
positions.  The work is in progress.

Key words:   Speech, articulators, topographical map, correlation, 
 multiple regression

Note: 
The above work was completed as part of the co-op Science and Engineering 
Research Semester at Los Alamos National Laboratory.  The author is an
undergraduate student at GSU and presently supported by the Office of Naval Research.