基于Android传感器的人体姿态识别方法  

Human activity recognition based on android sensors

作　　者：杨一涛 孙国梓[2] 王壮[2] YANG Yitao;SUN Guozi;WANG Zhuang(Department of Information Technology,Nanjing Forest Police College,Nanjing,China;School of Computer,Nanjing University of Posts and Telecommunications,Nanjing 210023,China)

机构地区：[1]南京森林警察学院信息技术学院 [2]南京邮电大学计算机学院,江苏南京210023

出　　处：《南昌大学学报（理科版）》2019年第6期616-620,共5页Journal of Nanchang University(Natural Science)

基　　金：中央高校基本科研业务费专项资金资助项目(LGZD201802);国家自然科学基金资助项目(61772284)。

摘　　要：当前基于各种传感器的Android设备已越来越多地应用于人们的生活中,提出一种基于三维加速度传感器的人体运动姿态转换的识别方法,该方法通过采集Android设备内嵌加速度传感器在人体运动过程中产生的加速度数据,提取不同运动姿态特征数据,进而识别走、跑、上楼和下楼等4种基本运动姿态,并通过持续判别法和前后续姿态判别法识别不同姿态间的转换。实验结果表明,针对单体的不同运动姿态识别率在90%左右,针对不同的行走-跑步间的姿态转换识别在93%左右,针对跌倒的检测在90%以上,均有较高的识别率。Current Android devices based on a variety of sensors have been increasingly used in people's lives.In this paper,a recognition method for different human motion conversion based on 3D acceleration sensors is proposed.It extracts the features of the different motion through gathering the acceleration data from the sensors in Android devices and recognizes four kinds of basic motion posture which are walking,running,go upstair and go downstair.Experimental results show that the monomer different motion posture recognition rate is 90%,the different walking-running posture conversion recognition is 93%,and the falling detection rate is over 90%.

关 键 词：穿戴设备 运动姿态识别 加速度传感器 跌倒检测 

分 类 号：TN915.08[电子电信—通信与信息系统]