“…Normally statistical recognizers are more reliable, however, they also have some disadvantages, it is very complex and computationally high to train and classify a large number of shapes because a large vocabulary is needed. The problems in fuzzy rules are lack of training and impossible to form exhaustive set of rules that can model all possibilities [26]. This paper describes a hybrid approach for the handwritten Urdu character recognition written in both Nasta'liq and Naskh styles.…”