On-Line Signature Verification by Explicit Solution to the Point Correspondence Problem

This paper presents a novel tool to be used for signature verification. The method is based on the vector context matrix concept that enables robust and precise extraction of the point correspondences between a sample and a database model. Using a digitizing tablet as a source for input, information of the pressure, speed and pen inclination is obtained in addition to the coordinates of the sampled stroke. The knowledge of the point correspondences enables studying these features pointwise and thereby increases the discriminatory power of verification systems. A simple system based on this tool has been tested on the samples of one signature from the SVC 2004 dataset with a perfect verification rate.

[1]  Ronny Martens,et al.  On-line signature verification by dynamic time-warping , 1996, Proceedings of 13th International Conference on Pattern Recognition.

[2]  C. Allgrove,et al.  Majority voting for improved signature verification , 2000 .

[3]  Marc Parizeau,et al.  Signature verification from position, velocity and acceleration signals: a comparative study , 1988, [1988 Proceedings] 9th International Conference on Pattern Recognition.

[4]  William J. Cook,et al.  Combinatorial optimization , 1997 .

[5]  Yutaka Fukui,et al.  On-line signature verification method utilizing feature extraction based on DWT , 2003, Proceedings of the 2003 International Symposium on Circuits and Systems, 2003. ISCAS '03..

[6]  W. Sardha Wijesoma,et al.  Automatic on-line signature verification based on multiple models , 2000, CIFEr.

[7]  Gerhard Rigoll,et al.  A systematic comparison between on-line and off-line methods for signature verification with hidden Markov models , 1998, Proceedings. Fourteenth International Conference on Pattern Recognition (Cat. No.98EX170).