Abstract

In this paper, a new on-line signature verification system using neural network is presented. The proposed system is based-on a newly developed Spatio-Temporal Artificial Neuron (SPAN), which is well adapted for the verification of Spatio-Temporal patterns. In this model, the strokes of a signature generated by a digitizing tablet is presented in form of a sequence of spikes corresponding to displacement of the stylus. The STAN has the capability to process continuous asynchronous Spatio-Temporal data sequence and compares them with the help Hermitian distance. The architecture of the proposed system is based on two modules : preprocessing and classification. The second module is based on neural architecture which has STANs as their neurons. This module is based on an adaptation of the RCE algorithm Our database includes 400 genuine signatures , 200 random forgery and 200 skilled forgery signatures that were collected from a population of 40 human subjects. Our signature database consist of the samples with about 100% size difference that are recognize thoroughly. Our verification system has achieved a false acceptance rate (FAR) of 7.5% and a false rejection rate (FRR) of 12.81%. Advantage of this method is using spiking neural network by Spatio-Temporal coding, using properly from signaturepsilas temporal feature, high speed of training and testing , using the less features in recognition and verification of signature, signature recognition in different size, the easy method with low expenses , not needing to any preprocessing such as rotating, transmit , normalization, filtering and no local limitation in digitizing tablet.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call