摘要
为减少因驾驶疲劳导致的交通安全事故,提出基于脑电（EEG）信号模糊熵（FE）的驾驶疲劳检测方法。开展在驾驶仿真模拟试验,采集28名被试模拟正常驾驶和疲劳驾驶的EEG信号;基于2种驾驶状态的EEG信号计算出FE值;运用随机森林（RF）、支持向量机（SVM）、决策树（DT）和K近邻（KNN）等4种分类器检测驾驶疲劳状态;利用多种性能指标及被试工作特征曲线（ROC）对驾驶疲劳检测结果进行分析比较。结果表明：基于疲劳驾驶状态下的EEG信号的FE值明显高于较正常驾驶状态下的值;4种分类器均可有效检测驾驶疲劳,其中K近邻的平均准确率达97.4%;基于EEG信号模糊熵的驾驶疲劳检测方法具有较好的鲁棒性和稳定性。
To prevent traffic accidents caused by driver fatigue,this study was aimed at developing a driving fatigue detection method based on EEG signal and fuzzy entropy. Firstly,EEG signals during simulated normal driving and simulated fatigue driving were acquired from 28 testees. Secondly,the FE values were calculated based on the EEG signals of two driving states. Then four classifiers（including random forest（RF）,support vector machine（SVM）,decision tree（DT） and K-nearest neighbor（KNN））were employed for detecting fatigue state. Finally,multiple performance indicators and ROC curve were adopted to analyze and compare the performance of driver fatigue detection. The results show that the FE value of fatigue driving based on EEG was significantly higher than that of normal driving,all four classifiers can detect driver fatigue state effectively,and K-nearest neighbor classifier achieves the optimal accuracy of 97.4%,and that the driver fatigue detection method based on fuzzy entropy of EEG signals has good robustness and stability.
作者
胡剑锋
王涛涛
HU Jianfeng;WANG Taotao(Center of Collaboration and Innovation, Jiangxi University of Technology, Nanchang Jiangxi 330098, Chin)
出处
《中国安全科学学报》
CAS
CSCD
北大核心
2018年第4期13-18,共6页
China Safety Science Journal
基金
国家自然科学基金资助(61762045)
江西省自然科学基金资助(20171BAB202031)
江西省教育厅科技项目重点课题(GJJ151146)
江西省博士后科研项目资助(2017KY33)
