基于HMM模型的老年人出行异常检测  

Detection of Abnormal Travel Behavior for the Old Based on Hidden Markov Model

作　　者：刘翔[1] 邹婷婷[1] 李仁旺[1] 王亚沛[1] 

机构地区：[1]浙江理工大学机械与自动控制学院,杭州310018

出　　处：《浙江理工大学学报（自然科学版）》2015年第3期410-413,共4页Journal of Zhejiang Sci-Tech University(Natural Sciences)

摘　　要：针对老年人出行异常的检测问题,提出一种基于HMM模型的出行行为检测方法。将出行轨迹的坐标点作为模型的训练数据,改进了HMM模型中隐藏状态的确定方法,建立老年人日常行为模型,然后提出一种判别分析方法用于找出检测轨迹对应的状态序列,计算该轨迹序列的输出概率并与阈值相比较,从而判断是否出现异常行为。通过实验证明了该方法对老年人行为异常检测有较高的准确性。This paper puts forward travel behavior detection method based on HMM model in allusion to abnormal travel behavior for the old. This paper improves hidden state determination method and sets up daily behavior mode of the old with coordinate points of travel track. Then, a discriminant analysis is proposed to find out stage series corresponding to detection track, calculate output probability of this track series and judge whether there is abnormal behavior through comparing with the threshold value. The experiment proves this method has high accuracy in detecting abnormal behavior of the old.

关 键 词：老年人 行为检测 聚类分析 HMM DBSCAN 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]