摘要
为了提高疲劳驾驶检测模型准确率和实时性,基于驾驶模拟实验,利用Smart Eye系统提取了驾驶人不同驾驶状态下眼动数据。基于眼动参数协议,提出了眨眼频率、PERCLOS、注视方向和注视时间4个特征参数的计算方法。分析了各特征参数的最优时窗,针对不同特征参数最优时窗差异,提出了滑移时窗的数据融合方法。基于支持向量机,搭建了疲劳驾驶检测模型。实验结果表明,该模型可以有效地进行疲劳状态检测,准确率能够达到83.84%。
In order to improve the accuracy and real time performance of the driver fatigue detection model,based on driving simulation experiment,the eye movement data in different driving states were collected using Smart Eye system. According to the protocol of Smart Eye system,a calculation method was proposed to obtain the characteristic parameters,including blink frequency,PERCOLS,gaze direction and fixation time. The best time window of different characteristic parameters was analyzed. For the best time window of each characteristic parameter was different,the slip time window was proposed to fuse the data. A diver fatigue detection model was developed based on the support vector machine. Validation tests showed that the method based on the driver's eye movements has a successful fatigue detection,whose accuracy reaches 83.84%.
出处
《哈尔滨工程大学学报》
EI
CAS
CSCD
北大核心
2015年第3期394-398,共5页
Journal of Harbin Engineering University
基金
教育部新世纪优秀人才基金资助项目(NCET-10-0435)
高校博士学科点专项科研基金资助项目(20110061110036)
吉林省人才开发基金资助项目(801121100417)
吉林省科技厅国际合作资助项目(20130413056GH)
关键词
疲劳驾驶
眼动特征
支持向量机
滑移时窗
时窗
检测模型
driver fatigue
eye movements
support vector machine
slip time window
time window
detection model