Analysis of Session Variability for Online Signature Verification Using Local Stability-Weighted DTW

摘要

In this study, we conduct the exhaustive experimental analysis of session variability to extract practical results related to a template-aging effect for a single-template strategy in online signature verification. We apply the novel single-template strategy that employs mean templates created by time-series averaging with Euclidean barycenter-based dynamic time warping (DTW) barycenter averaging and distance measurement using local stability-weighted DTW to improve the performance. Experimental results confirm the general effectiveness of the method for online signature verification using a public SUSIG Visual sub-corpus even when the session variability exists.