Signature Verification using Velocity-based Directional Filter Bank
Muhammad Hashim Khan, M. Aurangzeb Khan, Mohammad A. U. Khan, Sungyoung Lee · 2006
An on-line signature can be characterized with the help of its shape profile and dynamic characteristics. Shape and dynamics are both complementary aspects of a genuine signature. Their relationship should be exploited to develop a highly discriminative feature set for classification purpose. Conventionally these two characteristics are merely concatenated to form feature vector. However, a composite feature vector can be evolved while optimizing discriminative function satisfying shape and dynamics constraints simultaneously. One such simple strategy is suggested in this paper that employ 3D velocity selective directional filter banks. The output of the bank is a vector of directional energy with a given velocity constraint. Such a feature vector has been shown to be effective in developing an on-line signature verification system