Local Binary Pattern and Learning Vector Quantization for Classification of Principal Line of Palm-Hand

Abstract

Biometrics such as DNA, face, fingerprints, and iris still had disadvantages. The principal line of palm-hand biometric was expected to cover the weakness of the other biometric. This research was used dataset amounted to 150 images of palms-hand of the left-hand side. The dataset sourced 15 people who captured 10 times. The cropping technique that has used is the Region of Interest (ROI). Local Binary Pattern (LBP) was used to feature extraction. The feature extraction consists of the five parameters statistical. They were mean, variance, skewness, kurtosis, and entropy. Learning Vector Quantization (LVQ) was used to train the weight to produce optimal weight. The Confusion matrix method was used to evaluate the accuracy of the classification. The experiment was used the learning rates 0.01; 0.05; 0.1; 0.5; and 0.7. Based on testing and the experimental results, the highest accuracy obtained was on the learning rate value 0.5 which achieve 80%. In future work, we can explore with added the second-order statistics feature for better result.