CHARACTER RECOGNITION ALGORITHM ON THE BASE OF COMPETITIVE CELLULAR AUTOMATA

Abstract

This paper presents a new method for character recognition that is based on the concept of competing cellular automata. A new type of cellular automata, which move trajectory coincides with the character shape is represents. The advantage of this method is the insensitivity to the character size, lines thickness and proportion of fragments, distortion and partial overlapping symbols except the formation of joint lines. To optimize the recognition efficiency and speed offered the cellular automata competitive process; developed its algorithms and methods of interaction. To implement the proposed algorithms the modeling program was created. This software allowed to evaluate the effectiveness of cellular automata techniques and conduct experiments on English alphabet character recognition. It was demonstrated the successful recognition partly distorted characters and such imposed without forming joint lines. On the basis of these experiments authors concluded the prospects of using the proposed method in handwriting recognition. To create a real system it’s need to develop subsystem of interaction with scanning equipment, text segmentation principles, clearing it from the noise and automatic creation of cellular fields and output the recognition results.