Improved Critical Point Correspondence for On-line Signature Verification

an improved method to detect the critical points and obtain a matching path in 1 on-line signature verification. More critical points will be detected and the will be improved by the proposed method. We first detect critical points by ature curves for extremum points. We then introduce an improved DTW nal backward-merging DTW (BBMDTW), for the flexible matching of two 1D inally, the critical points and their correspondences will be regressed to 2D cause the 2D signature curve has more physical appearances of a signer. oint correspondence, DTW, Backward-merging, 2D regression. n-line signature verification can be generally classified into two categories: parameter-based [1]. Critical point detection and correspondence are two stages d signature verification system. They are very important and difficult, because it hat we detect consistent critical points from each signature produced by a same t topics in the field of on-line signature verification that how to detect more ly and how to make correspondence of segments between two signatures more e is divided into segments in its 2D curve by detecting 2D critical points such as points[2,3] or local maximum points of 2D local velocity and angular velocity TW algorithm is used for the flexible matching of two signatures[5,7,8, 9] Using ds, the number of critical points that are detected is not sufficient. That is to say, plex segments which should be divided further. Furthermore, using classical orrect segmentation cannot be overcome very well. vide a signature into more and simpler segments by 1D detection. Only 5 novel d for describing a 1D segment. An improved DTW algorithm, bidirectional