The New Jersey Institute of Technology's
Electronic Theses & Dissertations Project

Title:	Application of fuzzy theory to pattern recognition
Author:	Usechak, David
View Online:	njit-etd1983-019 (vii, 177 pages ~ 4.9 MB pdf)
Department:	Department of Electrical Engineering
Degree:	Doctor of Engineering Science
Program:	Electrical Engineering
Document Type:	Dissertation
Advisory Committee:	Reisman, Stanley S. (Committee chair) Blackmore, Denis L. (Committee member) Frank, Joseph (Committee member) Rosenstark, Sol (Committee member)
Date:	1983
Keywords:	Pattern recognition systems. Fuzzy algorithms.
Availability:	Unrestricted
Abstract:	Piecewise curve approximation is used to describe boundaries of objects in pictures and waveforms. The method consists of linear and quadratic piecewise polynomial approximations in which the error must not exceed a predetermined cost threshold. A fuzzy Bayes model is used to determine if a breakpoint exists within an interval In or between intervals In and In+1 and to determine whether these intervals can be merged for data compaction reasons. In order to achieve these objectives a new fast algorithm has been proposed which gives good curve/object fitting. This algorithm uses a technique for generating generalized inverse matrices once an initial generalized inverse matrix has been determined. The continuity requirements at the breakpoints are relaxed such that the only requirement is that the data point for interval In-1 is the starting point for interval In. Results of computer experiments with graphic outlines and radar data are reported.