A classifier approach to multi-screen switching based on low cost eye-trackers

During the past few years, research using eye movements has moved from controlled laboratory setup to the more complex natural task environments. In natural environments, the subjects will adopt to the dynamic environments, and this will involve head movements and various other pose interactions. Thus, in studies such as in visual perception, vigilance and fatigue the assessment paradigms should account for complex environments. The study presented in this paper addresses one such complex problem of using multiple low cost eye-trackers in a task using multiple computer screens. When using multiple screens in order to accurately record the eye tracking data the eye-trackers need to switch between eye-trackers depending on the head orientation with respect to the eye-trackers' field of view. Here, we present a classification-based switching mechanism for three eye-trackers. The proposed approach has given a higher accuracy of screen detection compared to the built-in switching mechanism of the eye-trackers under a natural task condition.