考虑生理特性的驾驶行为险态辨识研究

A Physiological Signal Based Method for Identifying Risk Status of Driving Behaviors

生理指标是驾驶过程中驾驶人状态最直观的体现,探讨生理指标与驾驶行为险态之间的关联关系对实现危险驾驶行为的辨识具有重要现实意义。以驾驶人生物反馈系统采集的3项生理特性指标为特征向量,采用皮尔逊(Pearson)相关系数法对不同生理指标与险态等级之间的关系进行深入分析,并在此基础上采用K-均值聚类方法构建驾驶行为险态辨识模型。通过对30组模拟驾驶实验数据的分析,最终得出驾驶人的血流量脉冲值(BVP)和皮肤表面电位(SC)与驾驶行为险态等级间存在显著正相关性(p<0.05),呼吸率(RESP)与驾驶行为险态间存在一定相关性,但是规律性不强。采用BVP和SC作为特征向量构建模型对驾驶行为险态辨识精度最高达到96%,对可忽略、可容忍和不可容忍3种状态的识别准确率分别达到97.33%,98.16%和88.16%。

Physiological indicators are the most intuitive reflection for statuses of drivers during driving process.Exploring relationships between physiological indices and risk driving behaviors has important practical significances in identification of driving risk.Three physiological indices collected by biofeedback system are regarded as feature vectors,and a method of Pearson correlation coefficient is used to analyze the relationships between different physiological indices and risk levels.Then a method of K-means cluster is used to develop an identification model.Data collected from 30 driving simulations is analyzed.The results show that blood volume pulse(BVP)and skin conductance(SC)of drivers are significantly influenced by different levels of risk driving behaviors(p<0.05),and there is no similar result in respiration(RESP).Recognition accuracy of the identification model which using BVP and SC as the feature vectors reaches to 96%,is the best.Recognition accuracy for different driving statuses of negligible,tolerable,and intolerable reaches 97.33%,98.16%,and 88.16%,respectively.