Personal palm vein identification using principal component analysis and probabilistic neural network

Biometrie is one type of physical characteristics that can be used for person identification as it has uniqueness. Palm vein is one of individual biometrics that attracts researcher attention recently. The advantages of palm vein compared to others are that palm vein represent if someone is still alive or dead, palm vein is hidden under the human skin, and palm vein is quite impossible to be imitated. However, a study related to the application of biometrie identification based on palm vein is still rarely done. Several methods and the implementations need to be observed, while the performances have to be proved as well. This study is aimed to apply Principal Component Analysis (PCA) as the feature extraction method and Probabilistic Neural Network (PNN) as the classification method. PCA is capable to reduce the feature dimension size, but still maintaining the important feature. So, the computational load is getting faster without damaging the extracted features. Meanwhile, PNN is a classification method that is known as a quick method in the training process but still optimizing the performance. The data set used in this research comes from Casia Database Multispectral. Some observations are done to obtain the optimal parameters setting. The highest accuracy obtained during the testing phase is 84% with the feature length used is 180 and the value of g is 0.001 for classifying 50 distinct individuals.