A Soft Computing Approach to Character Recognition

A character recognition system using soft computing techniques is presented in this paper. We define fuzzy sets on the Hough transform of each character pattern pixel and synthesize additional fuzzy sets by t-norms. The heights of these t-norms form an n-dimensional feature vector for the character. A 3n-dimensional vector is then generated from the n-dimensional feature vector by defining three linguistic fuzzy sets, namely, weak, moderate and strong for each feature element. These 3n-dimensional vectors for all the character patterns form a multilayer perceptron (MLP) input for training by the back propagation of errors. The fuzzy feature set is chosen after performing a sensitivity analysis of the multilayer perceptron outputs to the input features by a genetic algorithm. The MLP outputs also represent fuzzy sets denoting the belongingness of each input pattern to a number of fuzzy pattern classes. During recognition, outputs with high fuzzy set membership values are considered for a dictionary-based search to identify the ambiguous characters using word level knowledge. The system has been implemented for character recognition from printed English documents.