Offline Signature Verification Using Neural Network

Abstract

Even today an increasing number of transactions, especially financial, are being authorized via signatures, hence methods of automatic signature verification must be developed if authenticity is to be verified on a regular basis. Verification can be performed either Offline or Online based on the application. Online systems use dynamic information of a signature like velocity, acceleration and pressure captured at the time the signature is made. Offline systems work on the scanned image of a signature. In this paper we present a method for Offline Verification of signatures using a set of simple geometric features. The features that are used are Token length, Average values, Trend Coefficients and Standard Deviations of observation components. Before extracting the features, preprocessing of a scanned image is necessary to isolate the signature part and to remove any spurious noise present. The system is based on backpropagation neural network and is initially trained using a database of signatures obtained from the individual whose signatures have to be authenticated by the system. Then another set of test signatures of the same person are input to the system to check whether they are genuine or forgery. We either accept or reject the test signatures by using a suitable threshold. If the magnitude of the output of the neural network is less than a pre-defined threshold (corresponding to minimum acceptable degree of similarity), the test signature is verified to be genuine else detected as a forgery.