On-line signature verification based on three-dimensional force signals at critical points

It is an open problem to extract stable and distinguishable features from force signals of handwriting signatures. Based on the writing habits and kinematics of human, a hypothesis was proposed that force signal at critical points, e.g. the starting, the turning and the ending of a signature .etc, consists of personalized features which can distinguish signatures of different persons. Using stroke curvature, the critical points were extracted, and the 3-dimensional force signals at the critical points were also recorded. Employing Dynamic Time Warping (DTW) method, based on a preliminary signature database, verification experiments were carried out using the force signals of a whole stroke and the force signal at the critical points of the stroke respectively. The results look encouraging to verify the hypothesis.