Identification of humans using robust biometric features

Abstract

The size of the feature set is normally large in a recognition system using biometric data, such as Iris, face, fingerprints etc. As dimensionality reduction is an important problem in pattern recognition, it is necessary to reduce the dimensionality of the feature space for efficient biometric identification. In this paper, we present one of the major discriminative learning methods, namely, Direct Linear Discriminant Analysis (DLDA). Also, we specifically apply the multiresolution decomposition of 2-D discrete wavelet transform to extract the robust feature set of low dimensionality from the acquired biometric data and to decrease the complexity of computation when using DLDA. This method of features extraction is well suited to describe the shape of the biometric data while allowing the algorithm to be translation and rotation invariant. The Support Vector Machines (SVM) approach for comparing the similarity between the similar and different biometric data can be assessed to have the feature's discriminative power. In the experiments, we have showed that that the proposed method for human iris and face gave a efficient way of representing iris and face patterns. © Springer-Verlag 2004.