Multimodal Biometric Person Recognition by Feature Fusion

Fusion for multimodal biometrics can be done in various levels. Among them, the score and decision fusion levels have been widely studied in the literature, but feature fusion level is a relatively understudied problem. This paper proposes a feature fusion method based on the Simulated Annealing (SA) technique. A study is conducted to compare the results obtained by the SA technique with the ones obtained by the Genetic Algorithm (GA). The experimental result shows that both methods achieve the same accuracy performance in feature fusion level, while the SA technique is computationally more efficient than the GA method.