Vol. 30, Issue 2-3, pp. 317-333 (2000)

Abstract

A pattern recognition system with a computer-generated hologram used as feature extractor, useful in many applications, is considered in the paper. The method of optimization of such extractors, based on a modified evolutionary algorithm with elements of rough set theory used to define a proper objective function, is also presented. The theoretical investigations are supported with experimental works for two different applications of the optimized system. The applications are examples of a wide area of pattern recognition problems where proper characteristic features are not easy to determine. The results obtained confirm the enhancement of system optimized with proposed method compared to systems with standard computer-generated hologram.