摘要
针对已有基于信道状态信息(CSI)的行为识别方法存在冗余信息多、识别精度低等问题,提出一种基于CSI数据预处理的行为识别方法。首先通过计算子载波的贡献度进行子载波的选择,有效降低了CSI中的冗余信息;在此基础上,提出一种CSI动态特征增强算法,从原始CSI信息中分离出动态分量,实现对人体行为的准确表达,从而达到动态特征增强的目的。使用开源的CSI数据进行实验验证。结果表明:将预处理后的CSI数据用于行为识别,准确率较预处理前提升约30.5%;与WiFall、WiAnti预处理方案相比,本文所提方法准确率分别提高了7.5%与3.3%,证实了本文方法的有效性。
Aiming at the problems of redundant information and low recognition accuracy of existing CSI-based activity recognition methods,a behavior recognition approach based on CSI data preprocessing is introduced.Initially,subcarrier selection is performed based on the contribution of each subcarrier,effectively reducing redundancy in the CSI data.Building upon this,a CSI dynamic feature enhancement algorithm is proposed to extract dynamic components from raw CSI information,enabling accurate representation of human behaviors and thereby achieving dynamic feature enhancement.The open-source CSI behavioral data is experimentally validated,and the experimental results show that the accuracy of activity recognition after preprocessing is improved by about 30.5%.Compared with the preprocessing schemes of WiFall and WiAnti,the accuracy of the proposed method is improved by 7.5%and 3.3%,respectively,which proves its effectiveness..
作者
史伟光
姜皓元
SHI Weiguang;JIANG Haoyuan(School of Electronics and Information Engineering,Tiangong University,Tianjin 300387,China)
出处
《天津工业大学学报》
CAS
北大核心
2024年第6期66-72,共7页
Journal of Tiangong University
基金
天津市工程科技发展战略研究项目(22ZLGCGX0020)。
作者简介
通信作者:史伟光(1985—),男,博士,副教授,主要研究方向为射频定位、群体智能感知计算。E-mail:shiweiguang@tiangong.edu.cn。
