Mining generalized features for writer identification

This paper proposes generalized features of various handwriting in forensic documents for writer identification. In forensic documents, graphologies need to scrutinize, analyze and evaluate the features of suspected authors from questioned handwriting and compared these documents with the original handwriting. This is due to the uniqueness of the shape and style of handwriting that can be used for author's authentication. In this study, by acquiring the individuality features from these question documents will lead to the proposed concept of authorship invarianceness. However, this paper will focus on discretization concept that will probe authors' individuality representation by mining the features granularly. This is done by partitioning the attributes into writers' intervals. Our experiments have illustrated that the proposed discretization gives better identification rates compared to non-discretized features.