Detection of Gaze States: Fixation or Motion

Abstract

This paper addresses the problem of automated analysis of gaze states (fixation or motion). First, we present a novel dynamic and multi-level analysis structure for the detection of gaze states based on eye action units (AUs). Then the eye action units are analyzed both in spatial and temporal domain. And finally these eye AUs form the input to infer eye gaze state by employing Hidden Markov Models (HMMs) classifiers. The experimental results demonstrated reliable recognition of gaze states.