Multi-matcher dynamic signature recognition with protected and renewable templates

Abstract

In this paper we present a protected multi-matcher dynamic signature verification system which exploits score-level fusion techniques to combine Hidden Markov Model (HMM) and Dynamic Time Warping (DTW) classifiers. The considered on-line signature templates are treated with repeatable and non-invertible transformations, able to generate secure and renewable templates which can be fed to function-based matchers such as HMM and DTW. An extensive set of experiments shows that the combined use of HMM and DTW based classifiers guarantees remarkable performances in terms of both recognition rates and template renewability, while providing proper security to the employed biometrics.