• 首页
  • 期刊简介
  • 编委会
  • 投稿指南
  • 收录情况
  • 杂志订阅
  • 联系我们
引用本文:林志伟,刘梓隆,袁煜盛,倪沁玮,蔡志明.基于微多普勒特征的人体动作识别[J].软件工程,2024,27(3):21-25.【点击复制】
【打印本页】   【下载PDF全文】   【查看/发表评论】  【下载PDF阅读器】  
←前一篇|后一篇→ 过刊浏览
分享到: 微信 更多
基于微多普勒特征的人体动作识别
林志伟1, 刘梓隆1, 袁煜盛1, 倪沁玮1, 蔡志明1,2
(1.福建理工大学电子电气与物理学院, 福建 福州 350118;
2.福建理工大学电子信息与电气技术国家级实验教学示范中心, 福建 福州 350118)
2211905020@smail.fjut.edu.cn; 2211905025@smail.fjut.edu.cn; 2221908016@smail.fjut.edu.cn; 2221905029@smail.fjut.edu.cn; caizm@fjut.edu.cn
摘 要: 针对传统的视觉识别人体动作易受到光照变化、被遮挡和隐私问题等的影响,提出一种基于77 GHz毫米波雷达识别人体动作的方案。首先,使用77 GHz毫米波雷达采集人体动作的雷达回波样本,处理回波样本得到微多普勒频谱图;其次,首次提出一种新的自适应参数整流线性单元激活函数,将该激活函数与Resnet-18网络相结合;最后,将微多普勒频谱图数据集放入该网络训练并分类。实验结果表明,该方案对5种人体动作的平均识别准确率高达97.56%,较Resnet-PReLU网络提高了1.78%,有效地提高了对人体动作的识别精度。
关键词: 毫米波雷达;人体动作;微多普勒频谱图;激活函数
中图分类号: TP183    文献标识码: A
基金项目: 福建理工大学科研启动基金(GY-Z21064)
Human Motion Recognition Based on Micro-Doppler Features
LIN Zhiwei1, LIU Zilong1, YUAN Yusheng1, NI Qinwei1, CAI Zhiming1,2
(1. School of Electronic, Electrical Engineering and Physics, Fujian University of Technology, Fuzhou 350118, China;
2. National Demonstration Center f or Experimental Electronic Inf ormation and Electrical Technology Education, Fujian University of Technology, Fuzhou 350118, China)
2211905020@smail.fjut.edu.cn; 2211905025@smail.fjut.edu.cn; 2221908016@smail.fjut.edu.cn; 2221905029@smail.fjut.edu.cn; caizm@fjut.edu.cn
Abstract: This paper proposes a solution based on 77 GHz millimeter wave radar to address the impact of lighting changes, being occluded, and privacy issues on traditional visual recognition of human motions. Firstly, the 77 GHz millimeter wave radar is used to collect radar echo samples of human motions, and the echo samples are processed to obtain micro-Doppler spectrograms. Secondly, a new adaptive parameter rectification linear unit activation function that is proposed for the first time is combined with the Resnet-18 network. Finally, the micro-Doppler spectrogram dataset is placed into the network for training and classification. The experimental results show that the average recognition accuracy of this solution for five human moitions is as high as 97.56% , which is 1.78% higher than the Resnet-PReLU network, effectively improving the recognition accuracy of human motions.
Keywords: millimeter wave radar; human motions; micro-Doppler spectrogram; activation function


版权所有:软件工程杂志社
地址:辽宁省沈阳市浑南区新秀街2号 邮政编码:110179
电话:0411-84767887 传真:0411-84835089 Email:semagazine@neusoft.edu.cn
备案号:辽ICP备17007376号-1
技术支持:北京勤云科技发展有限公司

用微信扫一扫

用微信扫一扫