基于支持向量机的驾驶疲劳检测研究  

Detection of Driving Fatigue Based on Support Vector Machine

作　　者：李敏 李江天[1] 宋战兵[1] LI Min;LI Jiangtian;SONG Zhanbing(School of Automotive Engineering,Wuhan University of Technology,Wuhan 430070,China)

机构地区：[1]武汉理工大学汽车工程学院,湖北武汉430070

出　　处：《数字制造科学》2019年第2期131-135,共5页

摘　　要：为实现驾驶疲劳定量检测,设计了4 h高速单调路况模拟驾驶实验,通过编制的刺激信号采集反应时间,可穿戴设备采集皮电、肌电和心率信号,然后分析各信号变化规律,依据反应时间区分疲劳状态与清醒状态,选取生理信号特征指标,利用支持向量机(SVM)检测驾驶疲劳。研究发现:反应时间的变化呈跳跃性;皮电信号样本熵在疲劳状态下低于清醒状态,而大部分肌电信号样本熵在疲劳状态下高于清醒状态;基于SVM的驾驶疲劳检测算法准确率达86.25%,能有效识别驾驶疲劳。To realize the quantitative examination of driving fatigue,a simulated driving experiment of 4 hours high speed monotone road condition is designed.The signals in terms of GSR,EMG,and HR are respectively collected from skin,through wearable devices.Support vector machine(SVM)has been ultilise to detect the degree of driving fatigue by measuring the characteristic index of physiological signals,which is obtained by judging sober state or fatigue state from reaction time.It is found that the change of reaction time is hopping.The sample entropy of GSR in fatigue state is lower than that in the sober state,and the sample entropy of most EMG is higher in fatigue state than that in the sober state.Consequently,driving fatigue can be effectively identified based on SVM with the accuracy of 86.25%.

关 键 词：SVM 驾驶疲劳 皮电 肌电 样本熵 

分 类 号：U463.6[机械工程—车辆工程] TP181[交通运输工程—载运工具运用工程]