Pupillary response based cognitive workload index under luminance and emotional changes

Abstract

Pupillary response has been widely accepted as a physiological index of cognitive workload. It can be reliably measured with video-based eye trackers in a non-intrusive way. However, in practice commonly used measures such as pupil size or dilation might fail to evaluate cognitive workload due to various factors unrelated to workload, including luminance condition and emotional arousal. In this work, we investigate machine learning based feature extraction techniques that can both robustly index cognitive workload and adaptively handle changes of pupillary response caused by confounding factors unrelated to workload.