Recognition of Huffman codewords with a genetic-neural hybrid system

Abstract

Character classification is known to be one of many basic applications in the field of artificial neural networks (ANN), while data transmission with low size is important in the field of source coding. In this paper, we constructed an alphabet of 36 letters which are encoded with the Huffman algorithm and then classified with a back-propagation Feed Forward artificial neural network. Since an ANN is initialized with random weights, the performance is not always optimal. Therefore, we designed a simple genetic algorithm (SGA) that choses an ANN and optimizes its architecture to improve the recognition accuracy. The performance evaluation is given to show the effectiveness of the procedure used, where we reached an accuracy of 100%. © 2010 Springer-Verlag.