Decision Level Fusion Using t-Norms

Abstract

A multimodal biometric system employing the hand based modalities (i.e. palmprint, hand veins, and hand geometry) is developed. The proposed approach for the decision level fusion combines the decisions from these modalities using t-norms due to Hamacher, Yager, Weber, Schweizer and Sklar. These norms deal with the uncertainty and imperfection pervading the different sources of knowledge (error rates from different modalities). The proposed biometric system is quite computationally fast and outperforms the decision level fusion accomplished through the conventional rules (OR, AND) The experimental evaluation on a database of 100 users confirms the effectiveness of the decision level fusion. The preliminary results are encouraging in terms of the decision accuracy and computing efficiency.