Proposed Multi Feature Extraction Method for Off-line Arabic Handwriting Word Recognition

Abstract

This paper presents a proposed system for Arabic word recognition by using a set of techniques for feature extraction and KNN classifier. This system recognizes the Arabic word as one entity without segmentation approach .A proposed feature extraction method based on Discrete Wavelet Transform, upper/lower profile projection and the gradient (directional) feature. In order to evaluated the proposed system used IESK-arDB database and the experimental results showed the recognition rate of 89.05%.