Signature classification using optimum contour
J. Francisco Vargas, Carlos M. Travieso, Miguel A. Ferrer, Jesús B. Alonso · 2006
Abstract:- An offline handwritten signature classification approach is presented. The signature is parameterized using the radius values for contour points in the polar coordinates space. In order to determine the optimum outline for the classification task, nearest and faraway points were took into account. HMM and SVMLight model are analyzed in this work. Selecting faraway points offers better results, and using SVMLight model with polynomial kernel presents the best accuracy values obtained (99,02%).