Eye movement based information system indicates human behavior in virtual driving

Abstract

Humans modulate the behavior flexibly after timely receiving and processing information from the environment. To better understand and measure human behavior in the driving process, we integrate humans and the environment as a system. The eye-movement methodologies are used to provide a bridge between humans and environment. Thus, we conduct a goal-directed task in virtual driving to investigate the law of eye-movement that could characterize the humans (internal) and environmental (external) state measured by fixation distribution and optical flows distribution. The analysis of eye-movement data combined with the information-theoretic tool, transfer entropy, active information storage, quantify the humans’ cognitive effort and receiving information, and in fact, there is a balance (optimal) range between two, because of the mutual synergy and inhibition, whose quantified value is named balance of information processing. Subsequently, we update a system-level model, finding that those information measurements, transfer entropy, active information storage, and balance of information processing, all are included. This information set is information flow, which is quantified by the square root of Jensen-Shannon divergence (SRJSD), named information flow gain. What’s more, results also demonstrate that the influence of system-level information flow correlated with behavioral performance stronger than the separate measurements. In conclusion, we research humans’ eye-movement based on information theory to analyze behavioral performance. Besides driving, these measurements may be a predictor for other behaviors such as walking, running, etc. Still, the limitation is that the information flow may be a proxy of determinants of behavior.