Gaussian modeling and Discrete Cosine Transform for efficient and automatic palmprint identification

Abstract

Automatic personal identification using biometric information is playing a more and more important role in applications such as public security, access control, banking, etc. Palmprint identification is a subcategory of biometrics identification, which can efficiently used to identify the people. It is for this reason that palmprint-based identification is becoming increasingly popularity in recent years. In this paper, we present a novel scheme for palmprint identification using the multi-variate Gaussian Probability Density Function (GPDF) and two-dimensional Block based Discrete Cosine Transform (2D-BDCT). In this method, a palmprint is firstly divided into overlapping and equal-sized blocks, and then, applies the discrete cosine transform over each block. By using zigzag scan order (starting at the top-left) each transform block is reordered to produce the observation vector. Subsequently, we use the Gaussian probability density function for modeling the feature vector of each palmprint. Finally, Log-likelihood scores are used for palmprint matching. The proposed scheme is validated for their efficacy on PolyU palmprint database of 100 users. Our experimental results show the effectiveness and reliability of the proposed approach, which brings both high identification accuracy rate.