摘要
提出一种依据手机内置三维加速度传感器采集的人体日常行为数据来进行识别分类的方法。该方法对采集的原始加速度数据进行预处理,从水平和垂直方向提取多种统计特征,包括标准差、四分位差、信号幅度、偏度、峰度和相关系数等,由支持向量机分类器进行分类识别,可识别手机携带者站立、走路、跑步、上楼和下楼5种动作。通过对比分析实验结果,对不同实验者的平均识别正确率达到87.17%,验证了该方法的有效性。
A method of accurate analysis of basic activities with accelerometer data from a mobile phone is described in this paper.After sensor data are collected,preprocessing,vertical and horizontal feature extraction,classification steps are followed to build a training model.The system is trained to recognize five activities:staying,walking,running,ascending stairs,and descending stairs.Many statistical features are extracted such as standard deviation,signal magnitude areas,inter quartile range,skewness,kurtosis,entropy,and signal-pair correlation.For different subject,an average recognition accuracy of 87.17% is achieved.Results show the proposed method is effective.
出处
《西安邮电大学学报》
2014年第6期76-79,共4页
Journal of Xi’an University of Posts and Telecommunications
基金
国家自然科学基金资助项目(61373116)
关键词
人体行为识别
加速度传感器
支持向量机
human activity recognition
accelerometer data
support vector Machines(SVMs)