Towards optimal score level fusion for adaptive multi-biometric authentication system

Multi-biometric authentication systems have been extensively investigated due to their merit over unimodal systems. Multi-biometric systems not only overcome various security and privacy concerns of unimodal systems but also enhance overall performance & reliability. However, optimal fusion of multi-biometrics to overcome environmental challenges remains an open-ended research problem that was limitedly addressed. To address this issue, optimal score level fusion is proposed for a multi-biometric authentication system which includes biometrics namely fingerprint, finger-vein, and iris. The proposed optimal score level fusion technique exploits the Whale Optimization Algorithm (WOA) to efficiently weight belief assignment of finger-vein, iris, and fingerprint classifiers. The key features of WOA namely encircling, exploration, and exploitation are exploited to obtain optimal solutions for the fusion of multi-biometrics. For this, belief assignment for the score obtained from three modalities is computed using the Denoeux model. Furthermore, belief assignments are fused using the combination rule of PCR-6 (Proportional Conflict Redistribution) to achieve adaptiveness of the system to environmental effects. The proposed multi-biometric system is evaluated over a benchmark dataset namely, VERA finger vein, FVC 2002 and FVC 2006 fingerprint, and IITD PolyU iris. The experimental evaluation shows that the proposed technique attains, on an average, accuracy of 99.33% & equal error rate (EER) of 0.27% and these results are compared with the state-of-the-art system.