Comparative Analysis of Chameleon Swarm Optimization and Weighted Sum Fusion Techniques in Bi-Modal Recognition System

Bi-modal biometric systems integrate modalities such as palm-vein and face by fusion techniques to enhance biometric based security systems. Several techniques (especially evolutionary algorithms/swarm intelligence) have been developed and improvised as fusion techniques to reduce false positive rate and increase accuracies of biometric based recognition systems. However, these new techniques have not been adequately analyzed and compared with the conventional techniques like Weighted Sum rule. This study evaluates the performance of Chameleon Swarm Optimization (swarm intelligence algorithm) and Weighted Sum Rule as feature level fusion technique in a bi-modal recognition system. One thousand faces and palmveins samples were collected from a university environment. The acquired images were pre-processed to remove noisy areas and Local Binary Pattern was employed to extract features. The two outputs from face and palm-vein features were fused by the selected techniques. The fused features were subjected to classification by Support Vector Machine and the performance of these techniques was evaluated and compared. The results of the evaluation at an threshold of 0.85 showed that the Chameleon Swarm Optimization achieved a false positive rate (FPR) of 5.00% and accuracy of 95.67% and at a recognition time of 169.01µs while the Weighted Sum Rule achieved a FPR of 10.00%, and an accuracy of 92.33% at a recognition time of 116.35µs.