Fuzzy logic in the recognition of machine printed Arabic characters

This paper introduces a new recognition approach of machine printed Arabic characters. An input character is segmented into a main and a complement part. The features for classification are extracted from the number of complement strokes and the geometry of the main character and their projection profiles. The features are modeled by fuzzy linguistic values which provide a more expressive system for the characters. A set of fuzzy rules is used for classification. The system is tested on two fonts and the recognition rates achieved are very high.