疲劳特征的时序构建与分类.doc

中图法分类号: TP391.4 文献标志码: A 文章编号：（作者可不填） 论文引用格式： 精神疲劳实时监测的多面部特征时序分类模型 陈云华，张灵，丁伍洋，严明玉 （ 广东工业大学计算机学院计算机工程系， 广东广州 510006） 摘 要: 针对现有疲劳监测方法仅根据单帧图像嘴巴形态进行哈欠识别准确率低、采用阀值法分析眨眼参数适应性较差、无法对疲劳状态的过渡进行实时监测等问题，提出一种新的进行精神疲劳实时监测的多面部特征时序分类模型。首先，通过面部视觉特征提取得到张口度曲线与虹膜似圆比曲线；然后，采用滑动窗口分段、HMM建模等方法在张口度曲线的基础上构建哈欠特征时序及进行类别标记，在虹膜似圆比曲线的基础上构建眨眼持续时间时序及进行类别标记；最后，在HMM的基础上增加时间戳，以便自适应地选取时序初始时刻点和进行多个特征时序的同步与标记结果的融合。实验结果表明，本文模型可降低哈欠误判率，对不同年龄的人群眨眼具有很好的适应性，并可实现对精神疲劳过渡状态的实时监测。 关键词: 实时精神疲劳监测；虹膜似圆比；张口度；时间序列；滑动窗口法；隐马尔科夫模型 A classification model based on multiple facial feature time series for real-time mental fatigue monitoring Chen Yunhua, Zhang Ling, Ding Wuyang, Yan Mingyu (Department of Computer Engineering, Guangdong University of Technology, Guangzhou 510006, China) Abstract: In computer vision based fatigue monitoring, there are still some issues remained unresolved, including low recognition accuracy in yawn detection based on single-frame; poor adaptability in blink analysis through threshold; unable to monitor the transition stage of fatigue in real-time etc. Attempted to solve those problems, this paper proposed a new classification model based on two feature time series for real-time mental fatigue monitoring. First of all, calculate mouth opening degree and iris circularity ratio through facial visual feature extraction, and based on which we can generate corresponding time series called (the proportion of the time during which mouth openness exceeds the threshold ) time series and EBT (Eye Blink Time) time series. Then, using sliding window to partition and annotate the two kinds of time series and build HMM (Hidden Markov Model) for EBT time series. Finally, add a time stamp on HMM to adaptively calculate the initial time point of the next time series, in addition, we can use it to perform the synchronization and fusion of the two time series. Experimental results show that the promoted model can improve yawn detection rate, have good adaptability for blink features