Identification Based on Finger Knuckle Print (FKP) using Orthogonal Pseudo Zernike Moments

Researchers have recently found that skin pattern of the outer surface around the phalange which is called Finger Knuckle Print (FKP), is unique in different people and can be used as a distinct biometric characteristic. In this paper, a new identification system is proposed based on FKP images. In this system, Pseudo Zernike Moments (PZMs) which have the best overall performance in terms of robustness to noise, information redundancy and capability for image representation among the commonly used moments, are utilized for feature extraction. By using PZMs for the whole and partitioned FKP images, the global and local features which are concatenated to create the final feature vector are extracted. In the classification stage, a MLP neural network is utilized. Through considering recognition rate and equal error rate obtained for the proposed method in single instance (finger) and multi instance (finger) scenarios using PolyU FKP database, an acceptable improvement can be observed compared with the previous algorithms.