Offline handwritten Arabic writer identification using negative selection algorithm

Abstract

In pattern recognition; writer identification is one of the research are as that attract the researchers’ interest in the conduct of their studies. Writer’s identification of identity and its determination ability is not the only important thing to the writer, but the accuracy of this determination is considered as a big challenge. This study evaluates the accuracy of Arabic handwriting identification performance using the Bio-Inspired classifier. The study shows that the accuracy of the identification performance could be greatly improved with the Bio-Inspired classifier. The framework of the writer identification consists of three main phases: pre-processing phase, feature extraction phase, and classification phase. This research adopts IFN/ENIT Arabic Database which is constructed by Ecole National ed’Ingénieur de Tunis (ENIT) in Tunisia and Institute of Communications Technology in Germany (IFN). The images are enhanced by applying the threshold and conversion of the gray scale level images into black and white. Geometric Moment Function is used to extract the features from the images. Finally, the Bio-Inspired classifier is applied in this research with the use of Negative Selection Algorithm to classify and identify the writer. The obtained results show a promising ability of NSA in Writer identification. Other researchers could apply the NSA on handwriting languages that uses the same Arabic letters with different semantic such as Urdu as well as Farsi.