Computer Speech Processing for Speaker Identification

Abstract

A versatile speech processing system has been developed in the Department of Engineering Physics, Australian National University, using a PDP-15 computer with an image storage terminal. The application of spectrograms and cepstrograms to speaker identification has been explored and a comparative study in aural identification performed using a panel of untrained observers. The results of this work are presented and discussed in relation to the current scientific and legal status of speaker identification., with particular reference to the limitations of so-called "voiceprints".