Utilisation des caractéristiques statistiques et les Moments de Zernike pour la Reconnaissance des Lettres Arabes Manuscrits

Abstract

In this paper, we present an experimental study, which has for the objective of achieving better recognition rate of handwritten Arabic characters using statistical characteristics and moments of Zernike then a combination between the two. This work required the realization of a complete off-line handwritten Arabic character recognition system, the generation of a database of 4350 characters for different positions (beginning, middle, end and isolated) and the use of the SVM classification method due to the efficiency of this method. The preliminary results obtained are very encouraging and promising compared to the literature and show that the use of hybridization can be beneficial to better recognition of handwritten Arabic characters by the SVM method.