Swarm Intelligence Approaches for Score Level Fusion in Multimodal Biometric Authentication in Airport Security System

Abstract

Multimodal biometric system arise for overcoming the restrictions of unimodal  biometrics systems like non-universality, noisy data, spoofs, intra- and inter-class variabilities  and so on. In this research the modalities of finger print and iris are used as an Mutlimodal.  Matching score level fusion is typically favoured since matching scores are available with  ease and comprise enough data for distinguishing between genuine and fraudulent cases. For  a particular number of biometrics systems, matching scores may be generated for a  predetermined number of users with no knowledge of the underlying features extraction and  matching algorithms of all biometrics systems. Hence, combination of data comprised in the  matching scores appears both feasible and practicable. In this research work, similarity scores  are considered wherein matching scores are used for creating fused templates from the  feature of both finger print and Iris for identifying or verifying an individual’s identity. For  creating a template we propose an Swarm-Intelligene (SI) based algorithm Artificial Bee  Colony (ABC) with Artificial-Neural-Network (ANN) as an hybrid model (ABC-ANN). The  ANN refer to parallel distribution processors created by processing neurons features that  possess natural propensity for storage of experiential expertise and ensuring that it is  accessible for usage. The designs of ANNs owe their inspiration to the anatomy of the brain  which is a real world model of error-tolerant parallel processing that is both rapid and  powerful. ABC algorithm presumes the presence of a set of operations which resemble  certain features of the activity of honey bees. Fitness values to create a strong biometric  template refer to food source quality which is strongly linked to food location. The procedure  mimics bees’ search for precious food source giving rise to an analogous procedure for  discovering optimum solutions. The database templates and input data are compared by error  rates such as FRR, FAR, Accuracy-Rate, and Storage Space Complexity parameters. The  FAR, FRR, Accuracy, and Storage Space Complexity are compared with various threshold  levels. It had obtained minimal FRR, FAR and Storage Space Complexity for the proposed  ABC-ANN in experimental analysis and a higher Accuracy-Rate while comparing it with the  existing Gabor-HOG, AOFIS and ACNN method.