Holistic Approach for Urdu Character Recognition Using Modified Hmm

Automatic recognition of cursive handwritten script remains a challenging problem even with the promising improvement in classifier and computational power. Segmentation based approach for recognition of handwritten Urdu script has considerable computational overhead and has lower accuracy as compared to Roman and Chinese script due to additional segmentation error. Presence of complimentary characters in Urdu language makes it complicated as they have to be segmented into secondary strokes which are associated with a primary stroke. This first phase introduces a ligature based approach using Hidden Markov Model that provides solution for recognition of Urdu script. HMM database is divided into 54 subclasses based on the starting and ending shapes of the ligature. Twenty six time variant features have been selected for the base strokes. The sub division in classes reduces the time complexity and increases the efficiency. The second part of this chapter presents a segmentation free approach for recognition of Online Urdu handwritten script using hybrid classifier, HMM and Fuzzy logic. Trained data set consisting of HMM's for each stroke is classified into 62 sub