Offline handwritten Farsi cursive text recognition Using Hidden Markov Models

In this paper we address the problem ofrecognizing Farsi handwritten words. We extract two types offeatures from vertical stripes on word images: chain-code ofword boundary and distribution of foreground density across theimage word .The extracted feature vectors are coded using selforganizing vector quantization. The result codes are used fortraining the model of each word in the database. Each word ismodeled using discrete hidden Markov models (HMM).In order to evaluate the performance of the proposed system weconducted an experiment using new prepared database FARSA.We tested the proposed method using 198 word classes in thisdatabase. The result of experiment in compare with the existingmethods is very promising.

کلیدواژه ها:

Handwritten word recognition ، Hidden Markov model ، Self-organizing feature map ، FARSA database