Palm Vein Feature Extraction Method by Using Optimized DVH Local Binary Pattern

Intrinsic biometric, nowadays, has become a trend in research on human identification due to some disadvantages of the extrinsic biometric features. Extrinsic biometric features are easily imitated and lost as they are located outside the human body and are easy to change due to accidents. Therefore, in this paper we focus on a method which can extract a feature from an image of intrinsic biometric. Moreover, we use palm skin vein as the intrinsic biometric feature for human recognition application. The feature of an image can be extracted by using a specific method, such as Local binary pattern (LBP), which has been commonly used in many research works. A modified LBP, called cross-LBP (DVHLBP), has been proposed in our previous paper. DVHLBP has better performance compared with the conventional LBP. In this paper, we further optimize the DVHLBP method. In this paper, DVHLBP is used as the extraction feature algorithm on palm vein and histogram intersection is used for the matching process. In the simulation, the ratio of data model to data testing was 5:5. Testing was done by applying some scenarios. The optimization is done by examining the number of regions that yield the optimal threshold value. The optimal configuration is achieved when we use 8 neighborhood pixels with radius of 12, 16 regions. Simulation results show that the false accepted rate (FAR) and false rejected rate (FRR) are 0.01 and 0.01, respectively, with recognition rate of 99%. In addition, we show that the optimized DVHLBP has improvement in the accuracy and equal error rate (EER).