A Swarm intelligence approach for biometrics verification and identification

In this paper we investigate a swarm intelligence classification approach for both biometrics verification and identification problems. We model the problem by representing biometric templates as ants, grouped in colonies representing the clients of a biometrics authentication system. The biometric template classification process is modeled as the aggregation of ants to colonies. When test input data is captured -- a new ant in our representation -- it will be influenced by the deposited phermonones related to the population of the colonies. We experiment with the Aggregation Pheromone density based Classifier (APC), and our results show that APC outperforms ``traditional'' techniques -- like 1-nearest-neighbour and Support Vector Machines -- and we also show that performance of APC are comparable to several state of the art face verification algorithms. The results here presented let us conclude that swarm intelligence approaches represent a very promising direction for further investigations for biometrics verification and identification.