On-line signature verification based on dynamic segmentation and global and local matching

Abstract

We propose a method for signature verification based on natural-handwriting motion segmentation. Dynamic segmentation of the handwriting motion in accordance with motor commands is used to better correlate authentic signatures and to better reject forged ones. Utilization of personalized adaptive thresholds in signature segmentation and verification stages proves to produce good dissimilarity measures. Experiments show that 90% correct classification can be achieved.