Handwriting Arabic Character Recognition Using Features Combination

Abstract

The recognition of Arabic handwriting is a challenging problem to solve. The similarity among the fonts appears as a problem in the recognition processing. Various styles, shapes, and sizes which are personal and different across individuals make the Arabic handwriting recognition process even harder. In this paper, the data used are Arabic handwritten images with 101 sample characters, each of which is written by 15 different handwritten characters (total sample 101x15) with the same size (81x81 pixels). A well-chosen feature is crucial for making good recognition results. In this study, the researcher proposed a method of new features extraction to recognize Arabic handwriting. The features extraction was done by grabbing the value of similar features among various types of font writing, to be used as a new feature of the font. Then, City Block was used to compare the obtained feature to other features of the sample for classification. The Average accuracy value obtained in this study was up to 82%.