Discriminant Analysis and Naïve Bayes Classifier-Based Biometric Identification Using Finger Veins

Abstract

Finger vein identification is a dominating method of biometric technology used for authentication in a highly secure environment. Vein patterns are unique for each individual and it is underneath skin so there is less chance for forgery. In the current research work, finger vein features are extracted and verified for the purpose of authentication. The first step in this work is to pre-process the image obtained from the database. In order to get the region of interest (ROI) the threshold value is calculated using a standard deviation method followed by morphology-based functions available in the MATLAB software. After pre -processing a Gabor filter, fast filter, and freak descriptors are used. The features calculated at the freak descriptor processing are trained on classifiers namely discriminant and Naïve Bayes. The features trained to the classifiers are then fed again into the classifiers and cross verified to update the results of accuracy. The accuracy calculated using discriminant analysis is 94.46% and by using Naïve Bayes is 98.38%.