基于聚类和超限学习机分类模型的跌倒检测算法  

An Algorithm of Fall Detection Based on Clustering and Extreme Learning Machine Classification Model

作　　者：马宗方[1] 石晶 宋琳[1] 罗婵[1] 吴萌[1] MA Zongfang;SHI Jing;SONG Lin;LUO Chan;WU Meng(School of Information and Control Engineering,Xi'an University of Architecture and Technology,Xi'an 710055,China)

机构地区：[1]西安建筑科技大学信息与控制工程学院,陕西西安710055

出　　处：《济南大学学报（自然科学版）》2019年第1期50-55,共6页Journal of University of Jinan(Science and Technology)

基　　金：国家自然科学基金项目(61701388);住房和城乡建设部研究开发项目(K32016014);陕西省教育厅产业化项目(18JC017);西安建筑科技大学青年科技基金项目(QN1529;QN1630)

摘　　要：针对视频跌倒检测中对目标运动状态描述不完整和识别准确率不高的问题,提出一种基于聚类和超限学习机分类模型的跌倒检测算法,通过聚类算法对视频进行关键帧采样,建立特征模型,构建基于超限学习机的跌倒检测分类器,从而实现对跌倒行为的判断。结果表明:采用聚类的方法提取关键帧能够有效地剔除冗余图像帧,提高算法的准确度,也证明了超限学习机分类模型相较于其他分类器具有更好的分类性能。To deal with the inadequate description of target motion states and low accuracy of recognition in fall detection,a fall detection algorithm based on clustering and extreme learning machine classification model was proposed.Using the clustering algorithm,the key frames were extracted from videos,then the feature model was established.A algorithm based on extreme learning machine was built to judge the fall behavior.The results show that using the clustering method to extract key frames can effectively eliminate the redundant image frames and improve the algorithm accuracy,extreme learning machine has better classification performance than the other algorithms.

关 键 词：超限学习机 聚类 跌倒检测 运动姿态 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]