基于多帧间区域性光流特征的精神疲劳检测  

Mental fatigue detection based on multi-inter-domain optical flow characteristics

作　　者：姬秋敏 张灵[1] 罗源[1] 陈云华[1] 向文 JI Qiu-min, ZHANG Ling, LUO Yuan, CHEN Yun -hua, XIANG Wen(College of Computer, Guangdong University of Technology, Guangzhou 510000, Chin)

机构地区：[1]广东工业大学计算机学院,广东广州510000

出　　处：《计算机工程与设计》2018年第8期2582-2586,共5页Computer Engineering and Design

基　　金：广东省交通运输厅科技基金项目(科技-2016-02-030);广东省自然科学基金博士启动基金项目(2014A030310169);广东省自然科学基金面上基金项目(2016A030313703);广东省科技计划基金项目(2016B030305002);广东省自然科学基金项目(2016A030313713);广东省科技计划产学研基金项目(2017B090901005)

摘　　要：人体慢性疲劳表情具有持续时间长、帧间位移小的特点,使快速、准确的识别比较困难。分析上述问题,相比相位、能量、特征匹配的计算方法,基于梯度的光流方法计算简单且得到较好的结果,因此对基于梯度的全局光流技术进行研究,使用全局光流技术在连续多帧间局部区域内进行计算,获得稠密光流,得到相隔多帧的两幅图像的运动变化,解决持续时间长和帧间小位移的问题。通过计算眼部区域和嘴部区域的点运动情况提取光流特征,使用光流加速度模值,设置模值阈值参数提取开始打哈欠的那一帧,通过计算得出一段时间内眼睛闭合率和打哈欠频率。实验结果表明,该方法可以准确识别出疲劳表情。The expression of chronic fatigue for the human body has the characteristics of long duration and small inter-frame displacement,which makes it difficult to be identified quickly and accurately.According to the calculation method of phase,energy and feature matching,the gradient-based optical flow method is simple and better results are obtained.Therefore,the global optical flow technique based on gradient was studied using global optical flow technique.Dense optical flow was obtained,so that the motion of two images separated by multiple frames was obtained,which solved the problem of long duration and small interframe displacement.The characteristics of the optical flow were extracted according to the point motion of the eye region and the mouth region.According to the acceleration modulus value,the modulus threshold parameter was set to extract the frame that started yawning,and the eye closure rate and yawning frequency were calculated over a period of time.Experimental results show that the proposed method can accurately identify fatigue expressions.

关 键 词：疲劳表情 全局光流 光流特征 加速度 眼睛闭合率 哈欠频率 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]