Online handwriting Farsi character and number recognition based on hand movement direction using Hidden Markov Models

Abstract

Online handwriting recognition has many applications and the recognition with high accuracy is essential. In this paper, we introduce a method for online handwriting Farsi character and number recognition using Hidden Markov Models (HMM). First we recognize handwriting direction then we get some statistical and formatting features. The letters are classified by means of these features and then we use HMM for the recognition. We have some movements outside of the main body at the beginning or the end letters and number, so in this paper we put a Noise State at the beginning and end hmm and put an accepting state at the end to increase the recognition accuracy. We use Baum-Welch algorithm to introduce HMM and then give some samples. Note that the test results demonstrate the scalability of the proposed model and the recognition accuracy for numbers is 99.22% and for letters is 95.91%.