Rank Level Fusion of Multimodal Biometrics Based on Cross-Entropy Monte Carlo Method

Abstract

In unimodal biometric systems, there are several limitations like, non-universality, noisy data and other security risks. To overcome these, multimodal biometric systems are increasingly adopted. Multimodal biometric systems fuse information from multiple biometric traits. Rank level fusion is one of the approaches of information fusion for multimodal biometrics. In this paper, rank level fusion is considered as an optimization problem. Its aim is to minimize the distances between an aggregated rank list and each input rank list from individual biometric trait. A solution of this optimization problem has been proposed using cross-entropy (CE) Monte Carlo method. The proposed CE method uses two distance measures - namely, Spearman footrule and Kendall’s tau distances. Superiority of the proposed CE method based on above two distance measures over several existing rank level and score level fusion schemes is achieved on two different datasets.