Recognition of Printed Thai Characters Using Elliptic Fourier Descriptors and Genetic Neural Networks

Abstract

A recognition system for printed Thai character is proposed. We employ various methods i.e. Elliptic Fourier descriptors, Genetic Algorithms and Back-propagation Neural Networks to detecting boundary functions of characters, and attempt to normalize these functions by orientation and size variations of the symbols. Here, Genetic Algorithms and Neural Networks used to model the learning between the variables of boundary normalization of the characters. Finally, use of feed forward neural network to learn a boundary normalizes functions. A genetic algorithm is used to adjust the weights of the network interconnection. The experimental result shows the recognition rate are high though only a limited number of features have been involved. The result of this approach indicates that the systems can successfully recognized 3, 204 Thai characters with 94% accuracy.