Mental Workload Classification Using Electrocardiogram Data

Abstract

Electrocardiography offers a non-intrusive and portable alternative for real-time cognitive load assessment compared to traditional subjective measures. This study explores mental workload classification using machine learning models applied to electrocardiogram (ECG)-derived features. Data from a virtual flight task were processed and reduced to a binary classification problem, with feature standardization to address individual variability. Using nested leave-one-subject-out cross-validation, Random Forest, Support Vector Machine, Extreme Gradient Boosting, and Gradient Boosting models were evaluated. Random Forest achieved the highest F1 score (66.7%), indicating strong potential for ECG-based workload monitoring. These findings support the use of physiological signals for enhancing human-machine system safety and highlight future directions, including multimodal integration and deep learning approaches.