Mental workload classification using EEG in drivers: A personalized approach to occupational risk reduction

BackgroundMental state is a key factor influencing occupational health and safety, particularly in professional drivers who are exposed to high cognitive demands during long driving hours. Understanding how mental workload varies across individuals is essential for reducing risks related to fatigue and distraction in transportation occupations.ObjectiveThis study classified mental workload using EEG data collected during real-driving conditions and examined how personal factors (age, gender, and driving experience) contribute to inter-individual variability.MethodsEEG data were recorded from 39 active drivers during a real-world driving task on a route including main and secondary roads representing higher and lower workload conditions. Mental workload levels were labeled according to road type. A subject-specific classification approach was applied using four machine learning models, with individually optimized parameters.ResultsThe models discriminated between workload levels associated with different road types. The highest average accuracy reached 89.86%, with a maximum of 98.06%. The Multi-Layer Perceptron model combined with Principal Component Analysis gave the best results. Feature importance analysis revealed that frontal theta, beta, and particularly gamma band powers were the most discriminative features, although dominant features varied across individuals.ConclusionsThe findings demonstrate that EEG-based mental workload levels associated with different road types can be classified using subject-specific models. These results support the development of personalized and adaptive monitoring systems aimed at enhancing occupational road safety and driver health.