基于空间属性特征的毫米波雷达身体干扰识别  

Millimeter-wave radar body interference recognition based on spatial attribute features

作　　者：蔡嘉怡 初萍 庄伦涛 阳召成[1] CAI Jiayi;CHU Ping;ZHUANG Luntao;YANG Zhaocheng(College of Electronics and Information Engineering,Shenzhen University,Shenzhen 518000,China)

机构地区：[1]深圳大学电子与信息工程学院,广东深圳518000

出　　处：《系统工程与电子技术》2024年第10期3365-3374,共10页Systems Engineering and Electronics

基　　金：深圳市科技计划基础研究项目(JCYJ20190808142803565);广东省基础与基础应用研究项目(2022A1515140014)资助课题。

摘　　要：手势识别中,身体的移动容易被误判为手势动作,对手势识别造成干扰。因此,针对存在的身体干扰问题,提出了基于空间属性特征的身体干扰识别算法。在对毫米波雷达接收信号进行预处理后,首先分别对一维距离像和二维距离角度谱提取一维潜在目标和二维潜在目标,并对二维潜在目标进行连通域标记。然后,基于潜在目标及连通域提取出用于区分身体干扰和手势目标的空间属性特征。最后,采用支持向量机(support vector machine,SVM)分类器进行身体干扰识别。实验结果表明,所提方法能有效区分身体干扰和手势目标,单帧预测下准确率为97.3%,多帧预测下准确率为98.94%。In gesture recognition,the movement of the body is easily misjudged as a gesture action,which interferes the gesture recognition.Therefore,in view of the existing body interference problem,a body interference recognition algorithm based on spatial attribute features is proposed.After preprocessing the received signal of millimeter-wave radar,firstly,the one-dimensional potential targets and the two-dimensional potential targets are extracted from the one-dimensional range image and the two-dimensional range angle spectrum,and the connected domains are labeled for the two-dimensional potential targets.Then,based on the potential targets and the connected domains,the spatial attribute features used to distinguish between body interference and gesture targets are extracted.Finally,the support vector machine(SVM)classifier is used for body interference recognition.Experimental results show that the proposed method can effectively distinguish body interference from gesture targets,and the accuracy rate reaches 97.3%under single-frame prediction and 98.94%under multi-frame prediction.

关 键 词：毫米波雷达 手势识别 身体干扰 空间属性特征 特征提取 

分 类 号：TN959[电子电信—信号与信息处理]