An enhanced contextual DTW based system for online signature verification using Vector Quantization

This work presents an enhanced Dynamic Time Warping (DTW) based online signature verification system by utilizing the code-vectors generated from a Vector-Quantization (VQ) strategy. The DTW algorithms in the literature use only the distance score, obtained between the test signature and the genuine enrolled signatures, for the decision rule. The optimal warping path is constructed by placing constraints between the pairs of the sample points of the signatures, that are to be aligned. Hence, at times, sole dependence of the DTW scores may not be effective to discriminate the genuine and forgery signatures of an user, especially, when their values are very close. In order to alleviate this issue, we propose a novel scheme of scoring/voting the aligned pairs in the warping path by a set of code-vectors constructed from a VQ step. We subsequently fuse this score with that of the DTW, by popular score combination strategies, for verifying a test signature. As a second contribution, we consider the incorporation of contextual information in the formulation to reduce the equal error rate of the verification system. The experiments on the publicly available SVC 2004 and MCYT 100 databases confirm the efficacy of our proposal. To the best of our knowledge, this work is the first of its kind, that exploits the characteristics of the warping path for online signature verification. (C) 2016 Elsevier B.V. All rights reserved.