Cognitive state measurement from eye gaze analysis in an intelligent virtual reality driving system for autism intervention

Abstract

Autism Spectrum Disorder (ASD) is a group of neurodevelopmental disabilities with a high prevalence rate. While much research has focused on improving social communication deficits in ASD populations, less emphasis has been devoted to improving skills relevant for adult independent living, such as driving. In this paper, a novel virtual reality (VR)-based driving system with different difficulty levels of tasks is presented to train and improve driving skills of teenagers with ASD. The goal of this paper is to measure the cognitive load experienced by an individual with ASD while he is driving in the VR-based driving system. Several eye gaze features are identified that varied with cognitive load in an experiment participated by 12 teenagers with ASD. Several machine learning methods were compared and the ability of these methods to accurately measure cognitive load was validated with respect to the subjective rating of a therapist. Results will be used to build models in an intelligent VR-based driving system that can sense a participant's real-time cognitive load and offer driving tasks at an appropriate difficulty level in order to maximize the participant's long-term performance.