NC STATE UNIVERSITY
DEPT OF COMPUTER SCIENCE
COLLOQUIUM ANNOUNCEMENT

TITLE:  Disguised-voice speaker recognition

SPEAKER: Dr. Robert Rodman (NC State)
        
WHEN: May 6, Wednesday, 4:00pm

WHERE: 402A Withers
        
ABSTRACT: A program for carrying out research on the speaker recognition
of disguised voices is discussed. Such a program would consist of
the following:

1) Definition and classification of disguises
2) Creation of databases of disguised voices
3) Testing of conventional speaker recognition systems on the
disguised-voice database
4) The investigation of which basic methods of modeling the vocal tract 
 -- HMMs, vector quantization, neural nets, etc. -- are most effective on
the various types of disguised voices
5) The feasibility of automatically detecting the use of disguised voice
6) Methods for computationally "undisguising," or compensating for, a
disguised voice 
7) The use of several disguise methods at the same time