An Efficient Model for Isolated Vietnamese Handwritten Recognition

This paper proposes an efficient recognition model for isolated Vietnamese handwritten character recognition. Based on connected regions of image we try to determine character image that belongs to a group of characters which has either diacritical mark or not. In our model, character part and its diacritical mark part could be separately defined and they come independently for recognition. Hence, we apply SVM (support vector machines) classification where statistical features from a mark part and a character part are extracted independently. Finally, we join classified results together to have recognition outcome. Our test results over Vietnamese handwriting with 50,000 character samples collecting from 655 individuals show that the accuracy of our recognition model is over 90%.