Detecting mental states in drivers offers an opportunity to reduce accidents by triggering alerts and signaling the need for rest or renewed focus. Here we used electroencephalography (EEG) to measure brain signals in young drivers operating a driving simulator to detect mental states and predict accidents. We measured reaction times to unexpected hazardous events and correlated them with EEG signals measured from the frontal, parietal, and temporal cortices as well as the central sulcus (corresponding to motor cortex). We found that EEG signals in the relative beta (power in beta (13–30 Hz) relative to total power of the EEG (0.5–30 Hz)), alpha/delta, alpha/theta, beta/delta, beta/theta frequency bands were higher for collisions than successful collision avoidance, and that the key decision-making period is the 2nd second before braking. Importantly, a decision tree classifier trained on these neural signals predicted collision avoidance outcomes. Then based on random forest model, we extracted three critical neural signals (beta/delta_frontal, relative beta_parietal and relative beta_central Sulcus) to classify collision avoidance outcomes. Our findings suggest measuring EEG during driving may provide useful signals for enhancing driver safety.

通过触发警报并发出需要休息或重新集中注意力的信号，检测驾驶员的精神状态提供了减少事故的机会。在这里，我们使用脑电图 (EEG) 来测量操作驾驶模拟器的年轻驾驶员的大脑信号，以检测精神状态并预测事故。我们测量了对意外危险事件的反应时间，并将它们与从额叶、顶叶和颞叶皮质以及中央沟（对应于运动皮层）测量的 EEG 信号相关联。我们发现相对 Beta 中的 EEG 信号（beta 中的功率（13-30 Hz）相对于 EEG 的总功率（0.5-30 Hz））、alpha/delta、alpha/theta、beta/delta、beta/theta 频率碰撞的频带比成功避免碰撞的频带更高，并且关键的决策时间是制动前的第 2 秒。重要的是，在这些神经信号上训练的决策树分类器预测了避免碰撞的结果。然后基于随机森林模型，我们提取了三个关键神经信号（beta/delta_frontal、相对 beta_parietal 和相对 beta_central Sulcus）来对碰撞避免结果进行分类。我们的研究结果表明，在驾驶过程中测量脑电图可能为提高驾驶员安全提供有用的信号。