Low cost dynamic signature verification system

Abstract

Dynamic signature verification (DSV) makes use of parameters such as pen velocity, acceleration, jerk and pressure as described by R. Plamondon and G. Lorette (1989) and is an attractive biometric for identity verification since signature checking is already accepted by the public as noted by R. Carter (1990). The slow uptake of DSV by electronic fund transfer point of sale (EFTPOS) equipment manufacturers is partly due to the cost of the acquisition and processing hardware. The paper describes an artificial neural network (ANN) based system and a low cost digitising tablet for use with any non unique, non tethered pen. The tablet employs simple digital electronics and may be readily integrated into existing EFTPOS equipment. A low unit cost can be achieved using an ASIC mounted on a PCB with a screen printed polyester overlay film bonded using anisotropic conductive adhesive. It discriminates between pen and palm pressure and eliminates switching artefact. An ANN is used to fuse a number of broad features during the verification process. Preprocessing reduces the dimensionality of the data vectors used during enrolment and verification. The algorithms, which have produced excellent results with a small sample of users, leave the ANN to identify the salient information during training. These are simple enough to be implemented on the EFTPOS equipment's own microcontroller.