Statistical Approach to Person Identification via Unique Properties of the Oculomotor Plant
This paper presents a statistical approach that employs Hotelling's T-square test for the purpose of novel biometric identification based on unique static and dynamic properties of the oculomotor plant (OP) represented by mechanics of the eyeball, surrounding tissues, ligaments, and extraocular muscles. Proposed statistical approach yielded False Acceptance Rate of 0% and False Rejection Rate of 9.1%, providing a significant improvement against previously published study where identification was performed with k-nearest neighbor classification (KNN) and decision tree (C4.5) approaches with dynamic and static properties of the OP extracted via a horizontal linear homeomorphic mathematical model of OP from recorded eye the movement trace. In the current research two dimensional linear homeomorphic mathematical model was employed, allowing to extract OP biometric information from two dimensional saccades therefore providing more accurate identification rates. Current study involved 46 subjects with eye movement recordings done at 1000Hz.
biometrics, authentication, eye tracking, oculomotor plant, eye, Computer Science
Mechehoul, M., Jayarathna, S., & Komogortsev, O. (2010). Statistical approach to person identification via unique properties of the oculomotor plant (Report No. TXSTATE-CS-TR-2010-25). Texas State University-San Marcos, Department of Computer Science.