Recognition of handwritten tifinagh characters using gradient direction features

Optical Character Recognition system goal is to convert handwritten characters input images to an editable text. Many OCR techniques have been developed by researchers for Latin and Arabic languages. Amazigh language still have few works in this area. This paper emphasizes a new methodology to recognize the Tifinagh characters using zoning gradient features with a high accuracy and recognition rate. The new methodology is based on gradient direction as features and artificial neural networks as a classifier. The novelty of the new proposed system is the high accuracy and the training time which is very small compared with other classifiers. © 2005 – ongoing JATIT & LLS.