ON-LINE SIGNATURE TEMPLATES

An extensive set of experimental tests has to be performed to assess the effectiveness of a non-invertible transform-based biometric protection scheme. Specifically, the verification performances achievable with the proposed systems are discussed in Section V.A. Moreover, a renewability analysis is conducted in Section V.B.

The tests are conducted using the public version of the MCYT on-line signature corpus which comprises 100 users, for each of which 25 genuine signatures and 25 skilled forgeries have been captured during five different sessions. Forgers were asked to perform the imitations after observing the static images of the signatures to copy, having tried to mimic them at least 10 times, and by then writing the forgeries naturally without breaks or slowdowns. Specifically, the employed database is split into a training data set, comprising the first 30 users of the public MCYT database, and a test data set, comprising the remaining 70 users. The experimental results reported in the following, regarding both the recognition.

In this work an on-line signature-based biometric recognition system, where the original signature templates are transformed in a non-invertible way in order to provide protection and renewability, while guaranteeing verification performances similar to those of unprotected systems, has been proposed. Specifically, a baseline transform based on the use of convolutions, together with two extended versions of the baseline method, have been presented.

        Different kinds of template matching have been proposed and applied to the considered scenario, by exploiting the properties of HMM; DTW and score level fusion techniques.

        Several experimental tests, evaluated on the public MCYT signature database, have been performed to evaluate the effectiveness of the proposed approach, in terms of both recognition performances and renewability. Specifically, it has been observed that.