Classification of Arabic Autograph as Genuine ‎And Forged through a Combination of New ‎Attribute Extraction Techniques

Abstract

This study proposes a new framework for an Arabic autograph verification technique. It extracts certain dynamic attributes to distinguish between forged and genuine signatures. For this aim, this framework uses Adaptive Window Positioning to extract the uniqueness of signers in handwritten signatures and the specific characteristics of signers. Based on this framework, Arabic autograph are first divided into 14X14 windows; each fragment is wide enough to include sufficient information about signers’ styles and small enough to allow fast processing. Then, two types of fused attributes based on Discrete Cosine Transform and Discrete Wavelet Transform of region of interest have been proposed for attributes extraction. Finally, the Decision Tree is chosen to classify the autographs using the previous attributes as its input. The evaluations are carried out on the Arabic autograph. The results are very encouraging with verification rate 99.75% for sequential selection of forged and genuine autographs for Arabic autograph that significantly outperformed the most recent work in this field