Writer identification by means of explainable features: shapes of loops and lead-in strokes

Abstract

Writer identification is an important issue in forensic investigations of handwritten documents. A particularly well-established method employed by forensic experts is to (visually) explore distinguishing features of handwritten characters for comparing pieces of handwriting. Our research within the NWO Trigraph project aims at automizing this laborious process. In this paper, we propose a novel method for identifying a writer by means of features of loops and lead-in strokes of handwritten characters. Using a k-nearestneighbor classifier, we were able to yield a correct identification performance of 98% on a database of 41 writers. These results are promising and have great potential for use in the forensic practice, where the outcomes of handwritten document comparisons have to be justified via explainable features like the ones explored in this paper.