Abstract: At present, the determination of target action alarm threshold of driver condition monitoring is mostly based on technical ability and driver driving habits, which is lack of theoretical basis. Therefore, this study designs experiments, collects vehicle handling and running state parameters through driving simulator, and establishes vehicle running state evaluation index system. This paper proposes a method to cluster the driver's unsafe driving behavior data set by K-means algorithm to determine the alarm action threshold, which can provide a reference for the development of driver status monitoring system and the formulation of relevant test standards.

Key words: Driver condition monitoring; K-means algorithm; Driving simulator; Alarm threshold

摘要： 当前驾驶员状态监测目标动作报警阈值的确定多根据技术能力及驾驶员驾驶习惯讨论 得出，缺乏理论依据。为此，论文通过设计实验通过驾驶模拟器采集车辆操纵及运行状态参 数，建立车辆运行状态评价指标体系。提出一种通过 K-means 算法对驾驶人不安全驾驶行为 数据集进行聚类确定报警动作阈值的方法，可以为驾驶员状态监控系统的开发及相关测试标 准的制定提供参考。

关键词: 驾驶员状态监测；K-means 算法；驾驶模拟器；报警阈值