摘要
本研究提出了一种专门针对X波段电磁超表面雷达和机器学习技术的动态手势识别技术,其主要目标是通过结合单通道调频连续波(Frequency Modulated Continuous Wave,FMCW)体制收发机和可编程电磁超表面的波束扫描来实现手势识别和分类。在传统的雷达手势识别方法中,通常需要依赖于多输入多输出(Multiple-Input Multiple-Output,MIMO)雷达架构才能有效获取目标的角度信息。然而,MIMO雷达架构由于需要具备多链路,不仅会增加硬件成本,还会增加控制复杂度。而电磁超表面天线可以依据加载天线阵列的电压不同实现灵活的波束调控,利用这一特性获取到目标的角度信息。本研究基于自行搭建的X波段电磁超表面雷达,采用二维独立扫描,即包含一次水平扫描和一次垂直扫描,同时提出相应的雷达信号处理方法,从雷达回波数据中分别处理获得与传统FMCW MIMO雷达相似的角度-时间图(Angle-Time Map,ATM)以及距离-时间图(Range-Time Map,RTM),将特征图像输入机器学习进行分类识别。通过实验验证和定性定量分析,该方法可以较好的区分六种不同的动态手势动作,为电磁超表面雷达在手势动作识别的低成本应用拓展提供了新的思路。
This study introduces a dynamic hand gesture recognition technique specifically designed for X-band electromagnetic metasurface radar and machine learning technologies.The primary objective is to achieve gesture recognition and classification by integrating a single-channel Frequency Modulated Continuous Wave(FMCW)transceiver with a programmable electromagnetic metasurface for beam scanning.Traditional radar-based gesture recognition methods typically rely on Multiple-Input Multiple-Output(MIMO)radar architectures to effectively acquire the angular information of targets.However,MIMO radar systems,due to their requirement for multiple channels,not only increase hardware costs but also control complexity.In contrast,the electromagnetic metasurface antenna can flexibly control the beam by varying the voltage applied to the antenna array,thereby obtaining angular information of the targets.Based on a self-designed X-band electromagnetic metasurface radar,this research employs two-dimensional inde-pendent scanning,comprising one horizontal and one vertical scan,while proposing corresponding radar signal processing methods.These methods process the radar return data to generate Angle-Time Maps(ATMs)and Range-Time Maps(RTMs)similar to those obtained from traditional FMCW MIMO radars.These feature images are then fed into machine learning algorithms for classification and recognition.Through experimental validation and qualitative and quantitative analysis,this method demonstrates the a-bility to effectively distinguish six distinct dynamic hand gestures,offering a new perspective for the low-cost application expansion of electromagnetic metasurface radars in gesture recognition.
作者
潘伟
肖政
任忠芳
王海鹏
PAN Wei;XIAO Zheng;REN Zhong-fang;WANG Hai-peng(Research Center of Applied Electromagnetics,Nanjing University of Information Science and Technology,Nanjing 210044,China;State Key Laboratory of Millimeter Waves,Southeast University,Nanjing 210096,China)
出处
《中国电子科学研究院学报》
2025年第5期455-464,共10页
Journal of China Academy of Electronics and Information Technology
基金
国家自然科学基金项目(61801262)
东南大学毫米波国家重点实验室开放课题(K202312)
南京信息工程大学人才启动经费(2022r071)。
关键词
手势识别
机器学习
X波段超表面雷达
gesture recognition
machine learning
X-band metasurface radar
