An Effective Method for Multi-biometric Fusion using Simulated Annealing

An appropriate combination of multiple biometric sensors increases the reliability of verification through biometrics. In this paper we propose an effective method of fusion of biometrics based on a dynamic selection of threshold point of fingerprint and iris biometrics towards identifier of an optimal set of rules for fusion. The effectiveness of the method has been established using several benchmark databases using Simulated Annealing approach. The selection of a proper set of parameters for SA is a multi-objective decision making optimization problem. Initially the matching scores for individual biometric classifiers are computed. Next, a SA-based procedure is followed to simultaneously optimize the parameters and the fusion rules for fingerprint and iris biometrics. An experimental verification of the convergence nature of the simulated annealing method with the worst case behavior for optimum rule selection is analyzed and a comparative result of the method with the Ant colony optimization technique is also given.