Two stage classifierfor Arabic Handwritten Character Recognition

Abstract: In this paper we will present a two phase method for isolated Arabic handwritten character recognition system. The new method combines two levels based on two classifiers, a public and a private according to the similar features among characters.In the first level, we built a public classifier to deal with all character groups, each group contains characters with overlapped feature. The public classifier classifies the charactersin the SUST-ARG dataset (Sudan University for Sciences and Technology Arabic Recognition Group) to specified groups. In the second level, we created a private classifier for each group to recognize and classify the characters within a group.The system was applied to 34 Arabic characters and achieved 78.79% recognition rate for the tested dataset within the first level of the grouping model andachieved 93% recognition rate for the tested dataset using the two level models.

Keywords: Isolated Handwritten Arabic character recognition, Back Propagation, feature extraction, classifiers combination, Artificial Neural Network.