Arabic diacritics detection and fuzzy representation for segmented handwriting graphemes modeling

In this paper we present a new approach of Arabic diacritics modeling. The developed algorithm represents a section of the features extraction module of an online Arabic handwriting recognition system based on explicit grapheme segmentation strategy. The algorithm consists in three stages: first the detection of diacritics using the dimensions and the positions of the isolated handwriting strokes respect to the baseline. Then a fuzzy classification of the detected diacritics as simple dot, double merged dots, three merged dots or `shadda' using parameters representing their dimensions and shapes. Finally a diacritics fuzzy membership function is defined for each segmented main grapheme to calculate three summative rates of diacritics assignment for respectively `shadda', upper dots and lower dots diacritics.