Dynamic Signature Verification Using Embedded Sensors

Abstract

This paper presents a new method for signature verification using a pen equipped with sensors. Traditional dynamic signature verification methods use digitizing tablets to record data. Here real time data is gathered using sensors embedded in the pen as the person signs. These sensors capture dynamic information of the signing process such as instantaneous acceleration, rotation, and other data. After processing raw data, classification is made using a combination of techniques such as dynamic time warping and hidden Markov models with Gaussian mixtures. Along with global feature comparison this method yields low false acceptance rate and false rejection rate. Details of a prototype system and performance on human subjects are also presented.