Comparison of character recognition performance - bayes classifier and neural network methods

Abstract

In this project, two learning schemes had been tested and verified, in the recognition unit within the image processing environment of the Machine Vision Tool, namely the Bayes classifier method and the neural network-error back propagation gradient descent learning method. Their performances were analysed and compared. Sometimes the algorithms, especially the Bayes classifier method, fail to distinguish between two characters that are inherently similar, such as 'S" and '5', 'B' and '8'.