基于转向盘操作的疲劳驾驶检测方法  

A New Method for Detecting Driver Fatigue Using Steering Performance

作　　者：金立生[1] 李科勇[1] 牛清宁 高琳琳[1] 

机构地区：[1]吉林大学交通学院,长春130022 [2]公安部道路交通安全研究中心,北京100062

出　　处：《交通信息与安全》2014年第5期103-107,共5页Journal of Transport Information and Safety

基　　金：高校博士学科点专项科研基金项目(批准号:20110061110036);吉林省人才开发基金项目(批准号:801121100417);吉林省科技厅国际合作项目(批准号:20130413056GH);长春市科技局重大科技攻关计划项目(批准号:2013021-13KG05)资助

摘　　要：疲劳驾驶是引发道路交通事故的主要原因之一,研究利用转向盘操作行为特征检测驾驶人疲劳的方法对改善交通安全具有重要意义。研究通过分析基于驾驶模拟器的疲劳状态下的实验数据,提取了描述疲劳状态的不同特征参数,运用方差分析方法量化了不同驾驶状态下特征参数的差异性水平,优化出转向盘转角标准差、转向盘角速度标准差、转向盘转角变异系数、转向盘转角熵和零速百分比5个参数作为疲劳驾驶的有效特征参数组。建立了基于支持向量机的驾驶人疲劳状态检测模型,并采用测试集样本对搭建的模型进一步验证,结果表明该模型对驾驶人疲劳的模型检测准确率为81.33%,灵敏度为85.33%,特异度为77.33%。Driver fatigue is one of the major causes of road accidents. In this paper, a new detection method for driver fatigue based on steering performance is proposed. Experiments are conducted using a driving simulator to collect steering performance data at different fatigue levels. Five parameters, including standard deviation of steering angle, standard deviation of steering angle velocity, coefficient of variation of steering angle, steering angle entropy and percent- age of non-steering, are developed through the analysis of steering performance data under different fatigue levels. Analy- sis of variance is used to quantify whether these parameters are significantly different under different fatigue levels. These parameters are introduced as the input vector of a driving fatigue detection model that is developed based on support vector machine. The experimental results show that the detection accuracy of proposed model is 81.33%, sensitivity 95.33%, and specificity 77.33%.

关 键 词：交通安全 疲劳检测 驾驶人疲劳 转向盘操作 支持向量机 

分 类 号：U491[交通运输工程—交通运输规划与管理]