Several works in the recent literature on biometrics demonstrate the efficiency of the multimodal fusion to enhance performance and reliability of the automatic recognition. In this paper, we experimentally compare the behavior of different rules for integrating different biometric identification systems. We investigated how the benefits of the fusion change by varying the set of the fused modalities, the adopted fusion scheme and the performance of the individual matchers. The experiments were carried out on two multimodal databases, using face and fingerprint. We considered trained and fixed fusion methods at score, rank and decision level.
展开▼