Performance Evaluation of Feature Level Fusion for Multimodal Biometric Systems

Abstract

The biometrics authentication using multimodalities provides superior facilities over the traditional by means of identity verification due to their unique characteristics as biometric traits are not transferrable. The physiological (palmprint, signature, face, iris, etc.,) and behavioral (voice, key stroke dynamics, etc.,) characteristics of human beings are used to identify the individual. It plays an important role in several applications such as international border crossings, restricting physical access to places like nuclear plants or airports, controlling logical access to shared resources, remote financial transactions, distributing social welfare benefits etc. In this research work, an algorithm based on the fusion of face and iris modalities using Stationary Wavelet Transform (SWT) and Local Binary Pattern (LBP) is discussed. Firstly, the samples of face and iris are applied on SWT to extract the geometric and statistical features of face and iris. Next, the sample of low frequency (LL) sub band component of SWT is applied on LBP to extract the features. Principal Component Analysis (PCA) is used to reduce the dimensionality of each sample. The relevant characteristics of both face and iris are fused to create a patten for every individual. The obtained features are compared with the features of database images using Euclidean Distance classifier The ORL and CASIA iris datasets are used to evaluate the performance of the proposed model. The accuracy of 99.42% is recorded for the proposed model and showed robustness when compared with the other state-of-the-art method.