Minimizing Spatial Deformation Method for Online Signature Matching

Abstract

Elastic matching is an important algorithm for online signature verification. Most of elastic matching algorithms are based on dynamic time warping(DTW). In this paper, we propose a new points matching algorithm for online signature verification. After resampling a signature with equidistant spacing, we describe the resampling point by its (3+1) demesional((3+1)D) information: 3D spatial information (x-axis, y-axis, time) and a feature vector. A soft-assign matrix and two special restrictions are introduced for recording the correspondence between two signatures. With a deterministic annealing framework, the mapping between two sets of (3+1)D signature points is achieved by minimizing the warping energy of the thin-plate spline(TPS). Experimental results demonstrate that the proposed algorithm outperforms the DTW-based matching algorithms.