Improving Handwritten Signatures Fluency via the Lognormality Principle

Abstract

This chapter proposes two efficient methods to modify the fluency of dynamic signatures. The main idea is to modify the number of velocity minima or virtual target points and reconstruct the signature with the new virtual target points. If the number of virtual target points is reduced, the fluency is improved, and vice versa. The modification of the virtual target points is accomplished initially by linking the samples of an on-line signature by 8-connected Bresenham’s lines to obtain the interpolated trajectory. Then, the most perceptually important points are estimated from the corners in the trajectory. To this end, two methods are proposed. The first method, which we term resampling-wise, develops a lognormal synthetic velocity profile used for resampling the static trajectory. The second method, recovering-wise, consists in estimating the virtual target points from the perceptually important points in the trajectory, linking them by circular trajectories, and reconstructing the dynamic trajectory. Additionally, both methods can be used to generate synthetic on-line signatures from static trajectories. Both methods’ efficiency has been tested in automatic signature verification by increasing skilled forgeries’ fluidity with the proposed methods.