The Method Of Predicting Changes Of A Dynamic Signature Using Possibilities Of Population-Based Algorithms

Verification of a signature on the basis of its dynamics is an important issue of biometrics. This kind signature is called the dynamic signature. It can be represented, among others, by the set of features determined on the basis of time characteristics: pen velocity, pen pressure on the surface of a graphics tablet, etc. Values of the features can change over time, individually for each signer. Our previous research was related to the prediction of these changes to increase the effectiveness of a signature verification process. This approach was effective. The main purpose of this work is to compare the effectiveness of the methods for a prediction of signature features changes using selected population-based algorithms. They are used for learning of the fuzzy system used for prediction. Tests of the proposed approach were performed using ATVS-SLT DB database of the dynamic signatures.