A novel palmprint recognition system using patch based filter response

A novel palmprint recognition system is proposed in this paper. The proposed system consists of pre-processing, feature extraction and classification components. During preprocessing a region-of-interest for palm texture is extracted from a given hand image and for this thresholding and mathematical morphological based operators are employed. The features extracted from the detected palmprint texture is based on a method of ordering the codebook patterns obtained from patch based filter responses with a new way of feature representation scheme. Finally a multi-class linear SVM classifier is employed to do the classification job. The performance of the proposed palmprint recognition system is tested on four benchmark hand image databases, namely IITD, POLYU, CASIA and BOSPHORUS and this shows the efficacy of the proposed method.

1. Introduction

The biometric recognition system has great impact for providing security facilities in the area of banking applications, immigration sector, access control, e-commerce, border control systems and many mores. These systems claim the identity of a person (subject) based on their physiologi-calor behavioral characteristics such as iris, facial features, hand geometry, palmprint, voiceprint, fingerprints, gestures, gaits, handwritten signature and hand geometry of an individual. Among these traits hand biometric needs low-resolution camera or scanner for image acquisition. It is very less invasive and more easier to use in practice [7]. Moreover it is also stable over times for adults except some injuries and other deformation hazardous, the hand biometric is unsusceptible to major changes. Therefore, in this work we build a recognition system for hand biometric of the subjects.