Online Signature Verification using Deep Descriptors

Abstract

Signature verification is a technique used to counter signature forgery. In the past, the process began with staff at a bank, who is an expert, would confirm if a signature is genuine or forged. With the development of technology, now people no longer sign on paper, rather on a digital pad which can take more data which is recorded on paper, for example, pressure, azimuth and altitude angles. Examples of details captured from a digital pen include pen pressure, azimuth and altitude angles. This data is now used in various dynamic signature verification systems that achieve high accuracy on evaluation tests using different forms of artificial intelligence. This paper investigates using artificial intelligence in the form of a Convolutional Neural Network (CNN) followed by a Recurrent Neural Network (RNN) to verify signatures using the SVC 2004 and SigComp2009 online datasets and it achieved a testing accuracy of 97.05%.