Offline Arabic handwriting recognition system based on HMM

Great challenges are faced in the offline recognition of cursive Arabic handwriting. This paper presents a segmentation-free system based on Hidden Markov Model (HMM) to handle this problem, where character segmentation stage is avoided prior to recognition. The system first extracts a set of robust features on binary handwritten images by sliding windows. Then the proposed system builds character HMM models and learns word HMM models using embedded training. Finally, best word maximizing the a posteriori is located through Viterbi Algorithm. Experiments that have been implemented on the benchmark IFN/ENIT database show the average recognition rate of this system is 84.09%.