A fuzzy expert system for recognition of handwritten arabic sub-words

A fuzzy expert system for selected Arabic sub-words recognition is presented in this paper. For each sub-word pattern, membership values are determined for a number of fuzzy sets defined on the features extracted from the pattern. These sub-words consist of two characters and are written cursively, so, the first step is to segment the sub-words into two objects, main and secondary objects, keeping the two characters connected and making use of the general and special features in the recognition process. Presence or absence of dots in a sub-word and the number of such dots are fuzzy features, since the dot(s) may not appear exactly above or below the related character and they may appear mixed together. Closed loop is a fuzzy feature also. The proposed expert system consists of two main parts. First, the preprocessing part includes the feature extraction step that provides sufficient information to the inference engine. Second, the inference engine which applies the suitable set of fuzzy rules and aggregates them towards the final decision.