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

Area of Science:

• Neuroscience
• Occupational Health
• Human-Computer Interaction

Background:

• Mental workload significantly impacts occupational health and safety, especially for professional drivers facing high cognitive demands.
• Understanding individual differences in mental workload is crucial for mitigating fatigue and distraction risks in transportation.

Purpose of the Study:

• To classify mental workload using electroencephalography (EEG) data during real-world driving.
• To investigate the influence of personal factors (age, gender, driving experience) on inter-individual workload variability.

Main Methods:

• EEG data collected from 39 drivers during real-world driving tasks on varied road types (main vs. secondary).
• Subject-specific machine learning models (including Multi-Layer Perceptron with PCA) applied to classify workload levels.
• Analysis of feature importance, focusing on EEG frequency bands (theta, beta, gamma).

Main Results:

• Machine learning models successfully discriminated between workload levels associated with different road types, achieving up to 98.06% accuracy.
• The Multi-Layer Perceptron with Principal Component Analysis demonstrated superior performance.
• Frontal theta, beta, and gamma band powers were key discriminative features, though individual variations were noted.

Conclusions:

• EEG-based mental workload classification is feasible using subject-specific models in real-driving scenarios.
• Findings support the development of personalized monitoring systems for occupational road safety and driver well-being.