CATASTROPHE THEORY AND SOME DESIGN ASPECTS OF AUTOMATED IDENTIFICATION SYSTEMS (FROM CARNAHAN CONFERENCE ON CRIME COUNTERMEASURES - PROCEEDINGS, MAY 16-18, 1979, BY JOHN S JACKSON - SEE NCJ-62284)

• NCJ Number: 62308
• Author(s): D MACGOVERAN
• Date Published: 1979
• Length: 4 pages
• Corporate Author: Ores Publications
• Publication Format: Document
• Language: English
• Country: United States of America
• Note: *This document is currently unavailable from NCJRS.

Annotation

CATASTROPHE THEORY, A MATHEMATICAL METHOD, OF PREDICTING THE BEHAVIOR OF SYSTEMS UNDERGOING RAPID AND DISCONTINUOUS CHANGE, IS APPLIED TO THE DESIGN OF AUTOMATED IDENTIFICATION SYSTEMS.

Abstract

THE PROBLEM IN DESIGNING A PRACTICAL AUTOMATED IDENTIFICATION SYSTEM IS ESSENTIALLY ONE OF DETERMINING OPTIMUM CONDITIONS FOR AND LIMITATIONS INVOLVED IN PATTERN RECOGNITION. DIFFICULTIES IN CLUSTERING AND SEPARABILITY, ASSOCIATED WITH PATTERN RECOGNITION, APPEAR TO FALL INTO THE DOMAIN OF CATASTROPHE THEORY. AUTOMATED IDENTIFICATION SYSTEMS SHOULD BE DESIGNED TO CHOOSE FEATURES OR ADJUSTABLE PARAMETERS IN A WAY THAT REDUCES THE EXTENT TO WHICH IMAGES IN PATTERN RECOGNITION CLASSES OVERLAP IN THE FEATURE SPACE. ALTHOUGH NO GENERAL DECISION RULES ARE GIVEN BY CATASTROPHE THEORY, CHARACTERISTICS ARE SUGGESTED. A MINKOWSKIAN METRIC IS DESIRABLE, AS WELL AS THE INDEPENDENT TREATMENT OF PARAMETERS. AN EXAMPLE OF THIS PROCEDURE IS PRESENTED INVOLVING A SPECIFIC VOICE IDENTIFICATION SYSTEM. THE APPROACHES INCLUDE EXAMINATION OF CUE MATERIAL IN THE FORM OF AMPLITUDE-FREQUENCY-TIME MATRIXES AND EXTRACTION OF SPEAKER-DEPENDENT PARAMETERS FROM SPEECH SIGNALS WHICH ARE STATISTICALLY ANALYZED. FOR CRIMINOLOGICAL PURPOSES, AUTOMATED IDENTIFICATION SYSTEMS SHOULD BE DESIGNED WITH LIMITED POPULATIONS IN MIND AND SHOULD NEVER BE USED IN CASES WHERE POSITIVE IDENTIFICATION IS A REQUIREMENT UNLESS IT CAN BE SHOWN THAT POPULATIONS ARE SUFFICIENTLY LIMITED TO ALLOW WELL-CONTROLLED SYSTEM OPERATION. MATHEMATICAL EQUATIONS AND REFERENCES ARE INCLUDED. (DEP)