基于脑电信号的不良驾驶状态识别研究综述

Review on Recognition of Unsatisfactory Driving Status Based on Electroencephalogram

驾驶人是“人-车-路-环境”交通系统的重要环节之一,驾驶人状态与自身的驾驶行为密切相关。脑电信号(Electroencephalogram,EEG)作为驾驶人大脑活动的直接体现,能够有效地表征驾驶人当前时刻的驾驶状态。基于现有文献研究成果,阐述了EEG与驾驶人分心、疲劳及情绪等不良驾驶状态之间的内在关联,并对EEG研究中涉及的试验环境、数据处理及分析方法进行重点归纳总结。结果表明:大部分研究的本质可解释为不同驾驶人状态与EEG之间的定性与定量关系研究,通过志愿者模拟驾驶等方法收集EEG相关数据,利用线性或非线性等分析方法提取EEG特征值,以数学模型或神经网络模型对驾驶人的状态进行识别。进一步地,为提高状态识别模型的准确率,基于EEG的多源信息融合方法在不良驾驶状态等场景中的研究逐步增多,基于EEG在驾驶状态识别系统中的运用也逐步走向市场化。由此表明,目前基于EEG的驾驶人识别算法具有良好的安全应用潜力及前景,但在EEG特征提取、实时处理、不同驾驶人状态识别精度等方面仍具有较大改进空间。

Driver condition,as a crucial component of the“human-vehicle-road-environment”transport system,significantly affects driving behavior.Electroencephalogram(EEG)serves as a direct indicator of brain activity and can accurately reflect a driver's current state during driving.This paper begins by outlining the inherent relationship between EEG and adverse driving conditions,such as distraction,fatigue,and emotion based on the literature.Subsequently,key aspects of the test environment,data processing,and analysis methods employed in EEG research are summarized.The summary reveals that the essence of most studies can be interpreted as an exploration of the qualitative and quantitative relationships between various driver states and EEG.EEG data is collected through the simulated driving by volunteers,the EEG characteristic values are extracted by linear or nonlinear analysis methods,and then the driver's state is identified by mathematical models or neural network models.Furthermore,to enhance the accuracy of recognition models,research on multi-source information fusion based on EEG in scenarios like unsatisfactory driving state has gradually increased.The application of EEG in driving state recognition system is progressively moving towards commercialization.This indicates that current driving state recognition algorithms based on EEG possess promising safety application potential and prospects.Nonetheless,there remains significant room for improvement in areas,such as EEG feature extraction,real-time processing,and recognition accuracy across various driver states.