面向特殊人群行为识别的主动学习与预测方法  

Active learning and prediction method for activity recognition of special populations

作　　者：陈凡健[1] Chen Fanjian(Department of Computer Engineering,Maoming Vocational Technical College,Maoming 525000,China)

机构地区：[1]茂名职业技术学院计算机工程系,广东茂名525000

出　　处：《电子技术应用》2018年第11期116-120,共5页Application of Electronic Technique

摘　　要：老人等特殊人群的智能护理是人体行为识别研究的新方向。现有行为识别方法在学习样本时需要人工标记的样本数量多,在面向特殊人群行为识别应用时存在工作量大、识别率低的问题。为此,提出一种主动学习与预测方法,依据信息熵和互信息量构建目标函数,主动学习行为样本集,自动生成需要人工标记的样本子集。同时,以无向图模型作为行为描述子,依据信任传播方法进行类标签预测。实验结果表明,该方法需要人工标记的样本数量少,而且对特殊人群行为识别的识别率高。The intelligent nursing of the elderly and other special populations is the new research direction of human activity recognition.For the existing activity recognition methods,the number of samples that need to be manually marked when learning samples is large,and there is a problem that the workload is large and the recognition rate is low in the application of activity recognition for special populations.Therefore,an active learning and prediction method is proposed,which constructs the objective function according to the information entropy and mutual information,to learn activity sample set actively,and automatically generates the subset of the samples need to be manually marked.At the same time,the undirected graph model is used as the activity descriptor,and the class labels are predicted through belief propagation method.The experimental results show that this method requires a small number of manually labeled samples,and the recognition rate of activity recognition for special populations is high.

关 键 词：行为识别 行为预测 主动学习 无向图 智能护理 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]