Computer assisted voice identification system (CAVIS)
Hirotaka Nakasone, Craig Melvin · 2003
The major goal of the CAVIS project is to develop a system that is capable of dealing with transmission-independent and text-independent voice data, as well as rendering objective decisions. The need for such a system for law enforcement is discussed. The project's approach to handling different types of voice data, signal conditioning for cross channel frequency distortion, statistical processing for parameter extraction and selection, and speaker identification is described. Preliminary results of identification performance rates based on 21 male speakers are reported.>