An HMM-based online recognition system for Farsi handwritten words

In this paper, we propose a method for online Farsi handwritten words recognition. At first, words are broken to their sub-words. Each sub-word is made of some strokes. We assign a tag to each sub-word based on the positions and shapes of its sub-strokes. After that, we classify sub-words according to their tags. Some online features are extracted from the main-stroke after the preprocessing stage. Preprocessing contains operations such as dehooking, smoothing, normalization and boundary box equalization. Recognition process is consisted of some stages. First, the input word is divided into probable constructing sub-words, and recognition process is accomplished for each of them. For each sub-word, we find the class and sub-class of it, based on the tag and an extracted feature respectively. Some other features are extracted from the main-stroke of the sub-word, which are useful for training and testing the hidden Markov models as the classifier. These HMMs are the last level in our recognition system. In this paper, we use a 1000-sub-word database of the most frequently used Farsi words. The performance of the system in finding the classes and sub-classes of the sub-words is 99.45% and 99.91% respectively. The rate of correct performance of the HMMs is 82.96% making the total recognition rate of the system on the database 82.43%.