An Approach for Full Reinforcement-Based Biometric Score Fusion

Multibiometric systems have the potential to mitigate error rates and address certain inherent weaknesses found in unimodal systems. This study introduces an innovative scheme for user recognition in multibiometric systems, centered on a score-level fusion framework. The foundation of this framework lies in the full reinforcement operator (FRO), specifically estimating FRO through generator functions associated with triangular norms (T-norms and T-conorm). The efficiency of the proposed method has been showcased through an extensive set of experiments carried out on four commonly available benchmark databases: all three partitions of the National Institute of Standards and Technology (NIST) databases (Set 1, 2, 3), along with the XM2VTS database. Our method achieves superior accuracy compared to existing methods, reaching 100 % recognition on NIST-Set 1, 93.40 % on NIST-Set 2, and 94.54 % on the more challenging NIST-Set 3. The experimental findings illustrate that score fusion schemes based on FRO not only enhance verification rates when compared to current score-level fusion techniques (such as Asymmetric Aggregation Operators, Minimum, Maximum, T-norms, and Symmetric-Sum) but also offer a swift computational performance.