Hand Vein Recognition with Single-layer Feature Learning Model

Performance of feature extraction and representation, sticking point of image recognition task, will directly influence the accuracy of final recognition. The traditional feature extraction algorithm of vein recognition is based on the sufficient prior knowledge of analysis on vein information characteristics, the shortcoming of which reflects in long time consumption spent on tuning parameters and special selection about later classifier to guarantee the final recognition rate as high as possible. The paper makes the attempt to introduce the K-means model, single-layer feature representation architecture, to the vein recognition task with some targeted modification, and adopts the SVM at the link of classifiers design. Finally, the proposed approach is rigorously evaluated on the self-built database and achieves the state-of-the-art RR (Recognition Rate) of 98.34%, which demonstrates the effectiveness of the proposed model. Introduction Vein recognition, which is not introduced until 1990 by MacGregor and Welford [1], has become one of the most popular biometric identification methods with the advantage of unique, portable and inherent properties. The main characteristics we stress when putting vein recognition into practice not only because the vein patterns of individual is different even between identical twins [2], but also lies in the fact that it is easy-acceptable, anti-counterfeit and also with high recognition rate. Meeting all the requirements of biometric identification like other personal traits [3], vein recognition is also characterized with high convenience in image acquisition and feature representation which results in wide investigation on vein information research in hand covering palm vein [4-6], dorsal vein [7] [8] [9] and finger vein [10]. And the entire framework for the three kinds of recognition is more or less same, which covers vein image preprocessing, vein image feature extraction and representation, classifier design and vein recognition [11]. However the hand-crafted feature extraction methods of the traditional framework cannot ensure the high recognition accuracy, well robust performance, shorter time consumption to the same level. The analytical route cause is that the feature representation methods design are based on the single or mixture observation about the part of vein image information features, and it’s a common sense that the proposed methods cannot achieve the stage covering all the vein information representation based on the hand-crafted feature analytical result. The major goal in machine learning is to learn deep hierarchies of features for other tasks. A typical approach taken in the literature is to use an unsupervised learning algorithm to train a model of the unlabeled data and then use the results to extract interesting features from the data [12] [13] [14]. Depending on the choice of unsupervised learning scheme, it is sometimes difficult to make these systems work well. K-means has already been identified as a successful method to learn features from images by computer vision researchers. The popular “bag of features” model [15] [16] from the computer vision community is very similar to the pipeline that we will use in this chapter, and many conclusions here are similar to those identified by vision researchers [17] [18]. So the paper introduces the K-means to realize the vein feature representation system so as to find the feature distribution without adopting hand-crafted feature as the prior knowledge, and the experimental results demonstrate the efficiency of the proposed single-layer feature learning model in solving the feature learning problem in hand vein recognition task. 4th International Conference on Machinery, Materials and Information Technology Applications (ICMMITA 2016) Copyright © 2017, the Authors. Published by Atlantis Press. This is an open access article under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/). Advances in Computer Science Research, volume 71