A Brief Review of Classifiers used in OCR Applications

The performance of a recognition system depends upon the classifiers used for classification purpose. Powerful is the discrimination ability of a classifier, better is its recognition performance. The generalization ability of a classifier is measured on the basis of its performance in classifying the test patterns. There are various factors which affect generalization. Moreover, the feature extraction method(s) used for training a classifier also affects the performance of a classifier. In this paper, a brief theoretical review of various classifiers is made. The various characters of each are covered. The classifiers covered are Bayes, Parzen, probabilistic, polynomial, discriminant, radial basis networks, multi layer perceptron(MLP), k-NN, SVM and SOM.