Enhancing static biometric signature verification using Speeded-Up Robust Features

Abstract

Automatic biometric static signature verification performs a comparison between signature images (or preformed templates) to verify authenticity. Although widely recognised that performance enhancement can be achieved when using dynamic features, which use temporal/ constructional information, alongside static features, this scenario requires the capture of signatures using specialist sample equipment such a tablet device. The vast majority of (legacy) signatures across a range of important domains, including banking, legal and forensic applications, are in a static format. In this paper we use the Speeded-Up Robust Features (SURF) image registration technique in a novel application to static signature image matching. We use genuine and skilled forgery signatures from the GPDS960 dataset as test data and across a range of enrolment and SURF point distance configurations. The best performance from our method was 11.5% equal error rate by employing a product distance combination of 5 enrolment templates using the lowest 50% of returned registration-point distances. This encouraging result is in line with the current state-of-the-art performance.