Finger Vein Identification based on the Fusion of Nearest Neighbor and Sparse Representation based Classifiers

Objective: In this study, a new approach for personal identification using finger vein pattern is presented, to improve nearest neighbour algorithm by combining k-nearest neighbour and sparse representation based classifiers (KNN-SRC). Methods/Analysis: In the proposed KNN-SRC method, K numbers of best nearest neighbor samples were selected based on k NN classifier. Subsequently, the selected Ksamples were considered as the train samples for SRC classifier. Findings: Finger vein is a cutting-edge technology in biometrics that attracts attention of researchers from worldwide. As compared to the conventional biometric traits such as face, fingerprint and iris, finger vein is more secured and difficult to forge, as the veins are embedded in human tissue. Despite the intensive progress in feature extraction techniques from the captured vein images, there is a critical urge to develop an effective method for classification of the extracted features. Results of the present study using our own database revealed that, the proposed KNN-SRC method accomplished the analysis significantly faster than the SRC method which could be attributed to the reduced number of training samples. In addition, the KNN-SRC method gives higher accuracy than the common nearest neighbour (k NN) and SRC methods, individually, which could be attributed to sparsely representation of the test sample. Novelty /Improvement: In this method, recognition rate of finger vein images is improved by combining two classification techniques. For the first time, the KNN-SRC classification method was used for finger vein images.