摘要
为实现轻量级行为识别,弥补现有基于CSI(channel state information)行为识别方法中需要预先训练阶段和仅适用于单人场景的不足之处,对此提出一种新的行为识别方法(LBR)。该方法以室内信号传播模型中人运动速度与CSI频率间存在的关系为基础,将信号进行移动平均滤波和巴特沃斯低通滤波处理后,利用小波变换提取信号时频域特征并设计识别算法实现了行为识别。现实环境的实验结果表明,在损失少量准确率的情况下,LBR可实现一种无训练、两人场景下的行为识别,证明了该方法的轻便性与实用性。
In order to achieve lightweight behavior recognition,to make up for the shortcomings of existing behavioral identification methods based on channel state information that only fit to single person environment and need training stage,this paper proposed a new method of behavior recognition(LBR).This method based on the relationship between human velocity and CSI frequency in indoor signal propagation model,used wavelet transform to extract the time-frequency characteristics of the signal after undergoing moving average filtering and butterworth low pass filtering,then designed a recognition algorithm to obtain the behavior recognition results.The experiments of real environments show that LBR could achieve a training-free beha-vior recognition in double-person environment in the case of very little loss of accuracy,which proves the convenience and practicability of the method.
作者
姚青桦
崔然
Yao Qinghua;Cui Ran(College of Computer Science&Technology,Taiyuan University of Technology,Jinzhong Shanxi 030600,China)
出处
《计算机应用研究》
CSCD
北大核心
2018年第11期3397-3399,3404,共4页
Application Research of Computers
作者简介
姚青桦(1996-),男,山西晋城人,硕士研究生,主要研究方向为无线传感网络、物联网(timelessyqh@gmail.com);崔然(1994-),男,山西运城人,硕士研究生,主要研究方向为无线传感网络.
