Model-Based Estimation of Mental Workload in Drivers Using Pupil Size Measurements

Document Type

Conference Proceeding

Publication Date

1-1-2023

Publication Title

IEEE/ASME International Conference on Advanced Intelligent Mechatronics, AIM

Volume

2023-June

First Page

815

Keywords

and State-space model, Cognitive load detection, Expectation-Maximization algorithm, Eye-tracking, Human-computer interface, Kalman filter, Physiological signals, Pupil size

Last Page

821

Abstract

Passenger vehicles are increasingly adopting the use of automated driving systems (ADS) to help ease the workload of drivers and to improve road safety. These systems require human drivers to constantly maintain supervisory control of the ADS. For safe adoption and ADS, the attention or alertness of the driver needs to be continuously monitored. Past studies have demonstrated pupil dilation as an effective measure of cognitive load. However, the raw pupil data recorded using eye trackers are noisy which may result in poor classification of the cognitive load levels of the driver. In this paper, an approach to reduce the noise raw pupil size data obtained from eye trackers used by ADS is proposed. The proposed approach uses a Kalman filter to filter out high-frequency noise that arises due to sudden changes in ambient light, head/body movement, and measurement noise. Data collected from 16 participants were used to demonstrate the performance of the model-based pupil-size filtering approach presented in this paper. Results show an objective improvement in the potential to distinguish changes in pupil size due to various levels of cognitive workload experienced by participants.

DOI

10.1109/AIM46323.2023.10196230

ISBN

9781665476331

Link to Full Text

Share

COinS