AUTOMATIC SPEAKER RECOGNITION BY COMPUTERS (FROM CARNAHAN CONFERENCE ON CRIME COUNTERMEASURES, 1975 PROCEEDINGS BY JOHN S JACKSON - SEE NCJ-31252)

• NCJ Number: 31256
• Author(s): E BUNGE
• Date Published: 1975
• Length: 6 pages
• Corporate Author: University of Kentucky; Institute of Electrical and Electronics Engineers; Aerospace and Electronic Systems Society
• Publication Format: Document
• Language: English
• Country: United States of America

Annotation

THIS PAPER DESCRIBES THE PRINCIPLES OF AN AUTOMATICALLY WORKING SPEAKER RECOGNITION SYSTEM, AND OUTLINES THE BASIC DIFFICULTIES AND PITFALLS THAT MUST BE OVERCOME IN ORDER TO PRODUCE SUCH A SYSTEM.

Abstract

FOR SECURITY SYSTEMS IN BANKING AND LAW ENFORCEMENT, THE HUMAN VOICE MAY BE AN ESSENTIAL AID FOR VERIFYING A PERSON'S IDENTITY. A COMPUTER OPERATED SPEAKER RECOGNITION SYSTEM COULD ANALYZE VOICE PATTERNS OF THE SPEAKER, AND COMPARE THESE PATTERNS WITH THOSE STORED IN A COMPUTERIZED VOICE PATTERN LIBRARY. PROBLEMS IN PREPROCESSING THE SPEECH SIGNAL, 'TRAINING' THE SPEAKER IDENTIFICATION SYSTEM, AND TESTING THE RELIABILITY OF THE SYSTEM ARE DISCUSSED. IT IS STATED THAT A PROMISING APPROACH TO THESE PROBLEMS IS THE COMBINATION OF FAST SIGNAL PROCESSING TECHNIQUES AND TRAINABLE PATTERN RECOGNITION PROCEDURES. RESULTS OF A COMPARATIVE STUDY ARE DISCUSSED. (AUTHOR ABSTRACT MODIFIED)