Transfer Entropy Based Causality from Head Motion to Eye Movement for Visual Scanning in Virtual Driving

Abstract

AbstractVisual scanning is achieved by eye movement control for visual information acquisition and cognitive processing, which plays a critical role in undertaking common sensorimotor tasks such as driving. The specific coordination of the head and eyes, with head motions temporally preceding eye movements, is an important human behavior to make a key contribution to goal-directed visual scanning and sensorimotor driving. In this paper, we put forward a proposal of philosophy that this specific coordination of the head and eyes essentially indicates a unidirectional causality from head motion to eye movement. We propose to investigate transfer entropy for defining a quantitative measure of this unidirectional head-eye causality. A normalized version of the proposed causality measure is introduced for taking a role as an assessment proxy of driving. The plain transfer entropy-based definition has shown its statistical significance as the measure of causality and, the normalized version has demonstrated its good effectiveness for the evaluation of driving performance, with the verification in virtual reality-based psychophysical studies. This paper successfully suggests that the quantitative exploitation of causality based on the specific coordination of the head and eyes offers an effective approach to behaviometrics of visual scanning and sensorimotor activity.Author summaryThe coordination of head and eyes always exists in everyday sensorimotor driving tasks. Specifically, in goal-directed tasks, preparatory head motions guide eye movements to obtain and process relevant visual information for interacting with the surrounding environment. That is, the specific coordination of head and eyes involving head motions temporally preceding eye movement provides a mechanism for drivers to rely on prior knowledge for performing the tasks. As a matter of fact, this specific coordination of head and eyes essentially indicates, theoretically, a unidirectional causality from head motion to eye movement, leading to our proposal of causality philosophy. In this paper, an information-theoretic tool, transfer entropy, is exploited to capture the complex relationship between head motion and eye movement for obtaining the proposed measure of unidirectional causality. Furthermore, considering that the specific coordination of the head and eyes reflects the attention and cognitive state affecting the performance of sensorimotor tasks, we develop a normalized unidirectional causality measure as a proxy for the evaluation of driving performance. Psychophysical studies for goal-directed driving tasks are conducted based on virtual reality experimentation. Extensive results demonstrate a statistically significant correlation between the proposed normalized measure of causality and driving performance, which may provide a new and effective avenue for behaviometric applications. Practically, the merit of our proposed causality philosophy is that it is simple but effective, for obtaining an evaluation of the attentional and cognitive processes in driving tasks.