Estimation of mental workload using saccadic eye movements in a free-viewing task

Citations Over TimeTop 14% of 2011 papers

Abstract

This study proposes a new method to automatically estimate a person's mental workload (MWL) using a specific type of eye movements called saccadic intrusions (SI). Previously, the most accurate existing method to estimate MWL was the pupil diameter measure [1]. However, pupil diameter is not practical in a vehicle driving environment because it is overly sensitive to brightness changes. A new method should be independent from environment brightness changes, robust in most driving environments, and accurately reflect MWL. This study used SI as an indicator of MWL because eye movements, including SI, are independent from brightness changes. SI are a specific type of eye-gaze deviations. SI are known to be closely related to cognitive activities [2], [3]. This means that SI may be also closely related to MWL. Eye movements were recorded using a non-intrusive eye tracking camera, located 550 mm away from a participant. Participants were instructed to move their eye gaze to examine a highway driving scenery picture. In the data set of the recorded eye movements, our new algorithm detected SI and quantified SI behavior into a SI measure. Participants were also engaged in a secondary N-back task. The N-back task is a popular task used in cognitive sciences to systematically control a MWL level of participants. In our results, all 14 participants exhibited more SI eye movements when their MWL level was high compared to when their MWL level was low. Moreover, our results showed that the SI measure was a more accurate measure of MWL than the pupil diameter measure. This finding indicates that MWL of the person can be estimated by observation of SI eye movements. This new method has a wide range of applications. One of them is to predict a person's MWL, thus predicting when a person is capable of driving a vehicle in a safe or dangerous manner.

Related Papers

• → Passive driver gaze tracking with active appearance models(2018)164 cited
• → What Are You Looking at?(2011)24 cited
• → Gaze Estimation Based on Single Camera(2013)2 cited
• Improving Visual Gaze Estimation by Saliency(2012)
• Eye Tracking Gaze Estimation Based on Marker Detection(2011)