Bengali character recognition using genetic algorithm

Abstract

Genetic Algorithms have been quite successfully applied to a number of problems. The most outstanding results come from the field of parameter optimization, where the coding is rather straightforward. A few attempts were made to its application to pattern recognition in the framework of supervised learning technique. In this endeavour, A character recognition system using Genetic Algorithm has been developed. The system is intended to recognize printed Bengali characters. The model proposed for this system consists of a preprocessor followed by a Genetic Algorithm classifier. At preprocessing phase, projection from each active bit of a pattern has been scaled and translated to fit a standard size. The second part of the system comprises a Genetic Algorithm classifier which generates a set of rules based on the extracted features of the patterns. The rules are generated in such a way that only the distinctive features of a pattern are reflected in the rule. After being trained using a training set of character patterns, the system has been able to classifY test character patterns correctly. The proposed model has been tested with two complete character sets of Bengali alphabet and rigorous experiments have been carried out to see how the performance of Genetic Algorithm as a classifier varies at different parameter settings in the context of Bengali character recognition.