Fusion of Iris and Palmprint for Multimodal Biometric Authentication

This paper presents a multimodal biometric system for authentication, based on the fusion of iris and palmprint. We propose an approach for feature extraction of each modality by using wavelet packet decomposition at four levels. This gives 256 packets which can generate a compact binary code. It is obtained from the first three highest energy peaks to compute an adapted threshold that enable to affect 0 or 1 to each wavelet packet. Different fusion strategies were tested at different levels: feature level, score level and error level. The first fusion is a simple concatenation of iris and palmprint codes. The second employs a weighted sum rule to matching scores. The third applies the Hamacher t-norm to the errors. The proposed approach and each fusion strategy were tested for their accuracy on the Casia iris database fused with the Casia palmprint database, and then with the PolyU database. The proposed approach for multimodal biometric system achieves a recognition improvement with each fusion method.