Authors: Youming Zhang; Martti Juhola
Addresses: School of Information Sciences, University of Tampere, Kalevantie 4, 33014 Tampere, Finland ' School of Information Sciences, University of Tampere, Kalevantie 4, 33014 Tampere, Finland
Abstract: The biometric verification of users of computers or other machines is usually performed with fingerprints, face, iris or palm images. Eye movements have seldom been studied for biometric verification, although in the future their use will perhaps extend from laboratory applications to integrated parts of computer interfaces. Eye movements have long been studied in medical and psychological applications. We noticed that there are differences between saccade eye movements of individuals, even in a group of young people approximately of the same age. We measured saccades from 68 subjects by performing the same stimulation for each to obtain comparable data. We tested two verification conditions: a) an authenticated user vs. all other subjects and b) an impostor vs. an authenticated user and others. Randomised classifications with discriminant analysis, k-d tree and k nearest neighbour searching, decision trees and the naïve Bayesian rule were run, but logistic discriminant analysis produced the best results.
Keywords: eye movements; saccade; user verification; signal analysis; classification; biometric verification; user authentication; biometrics; logistic discriminant analysis.
International Journal of Biometrics, 2014 Vol.6 No.2, pp.106 - 124
Received: 20 Jun 2012
Accepted: 24 Feb 2013
Published online: 23 May 2014 *