基于线性回归分析的重复动作计数估计  

Repetitive Action Counting Based on Linear Regression Analysis

作　　者：吴艳春 孙红卫[2] 尹建芹[3] WU Yanchun;SUN Hongwei;YIN Jianqin(School of Information Science and Engineering,University of Jinan,Jinan 250022,China;School of Mathematical Sciences,University of Jinan,Jinan 250022,China;School of Automation,Beijing University of Posts and Telecommunications,Beijing 100876,China)

机构地区：[1]济南大学信息科学与工程学院,山东济南250022 [2]济南大学数学科学学院,山东济南250022 [3]北京邮电大学自动化学院,北京100876

出　　处：《济南大学学报（自然科学版）》2019年第6期496-499,共4页Journal of University of Jinan(Science and Technology)

基　　金：国家自然科学基金项目(11671171)

摘　　要：针对复杂的动态场景中平稳重复运动计数准确率较低的问题,提出基于线性回归分析方法对特定重复动作进行计数估计,根据同一动作在运动频率和运动形态上的相似性,对动作执行时间和重复次数之间的相关关系进行建模分析,在自建的数据集中进行实验,对视频重复动作计数进行预测和评估。结果表明:该方法不受复杂现实场景的干扰,对多样性的动作特征不敏感,且学习参数较少,简单快速;在自建的数据集中测试平均绝对错误率为12.94%,说明该方法对重复动作计数是有效的。To solve the problem of low accuracy of repetitive action counting for stationary repetitive motions in complex dynamic scenes,a method based on linear regression analysis was proposed to estimate the specific repetitive actions.Due to the similarity of the motion frequency and motion mode for the same action,the correlation between the execution time of the action and the number of repetitions was modeled and analyzed,and the experiment was carried out in the self-built data set to predict and evaluate the video repeated action count.The results show that this method is not disturbed by complex realistic scenes and is insensitive to diverse motion features,and it has fewer learning parameters and is simple and fast.The mean absolute error rate is 12.94% in the self-built data set,which indicates the effectiveness of the method.

关 键 词：线性回归分析 重复动作计数 数理统计 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]