Recognition of drivers' mental workload in simulated driving using EEG signals from different brain regions and machine learning algorithms

Accurate recognition of drivers' mental workload is essential to reduce traffic accidents caused by mental overload. However, it remains challenging to effectively and reliably recognize mental workload, especially with EEG signals. This study aimed to develop EEG-based mental workload recognition models in simulated driving and explore the specific roles of brain regions in mental workload recognition. Simulated driving task experiments were designed based on typical driving scenarios, leveraging the correlation between mental workload and EEG responses. EEG data from 24 drivers were collected. EEG features were extracted from four brain regions and applied to four typical machine learning algorithms to develop driver mental workload recognition models. The results showed that, δ rhythm in the Frontal lobe and the θ and α rhythms in the Temporal lobe are sensitive to mental workload across different driving task levels. While all the four brain regions achieved moderate to high model accuracy (ranged from 83.29% to 87.80%), frontal lobe yielded the highest classification accuracy of up to 87.80% with Random Forest algorithm. Our findings offer valuable insights for the use of EEG signals to recognize drivers’ mental workload, thereby contributing to improved driving safety.