Multi-class SVM classifier for english handwritten digit recognition using manual class segmentation

Abstract

A new method for recognition of isolated handwritten English digits is presented here. This method is based on Support Vector Machines (SVMs). Mean and standard deviation of each digit is considered as the features. Using these features, multiple SVM classifiers are trained to separate different classes of digits. Support vector machine are based on the concept of decision planes that defines the decision boundaries. The decision plane is one that separates between the set of digits having different class membership. The approach works in four steps 1) Preprocessing 2) Feature extraction 3) Classification 4) detection. A database of 100 different representation of each digit is constructed for the training database. The digits are first manually segmented into 5 classes to minimize the time required to obtain the hyperplane. Then the input is again check against the two classes by 2-class SVM classifier. Experiments show that the proposed features can provide a very good recognition result using Support Vector Machines at a recognition rate 97%, compared with 91.25% obtained by MLP neural network classifier using the same features and test set. Copyright 2009 ACM.