基于超表面的智能电磁感知:理论、系统与实验  

Metasurface-assisted intelligent electromagnetic sensing:theory,design and experiment

作　　者：双雅 李力 王卓[1] 魏梦麟 李廉林[1] SHUANG Ya;LI Li;WANG Zhuo;WEI Menglin;LI Lianlin(State Key Laboratory of Advanced Optical Communication Systems and Networks,Department of Electronics,Peking University,Beijing 100871,China;Beijing Aerospace Measurement&Control Technology Co.Ltd.,Beijing 100871,China)

机构地区：[1]北京大学信息科学技术学院,北京100871 [2]北京航天测控技术有限公司,北京100871

出　　处：《电波科学学报》2021年第6期858-866,共9页Chinese Journal of Radio Science

基　　金：国家重点研究研发项目(2017YFA0700203)。

摘　　要：电磁感知是现代社会迫切需要的非接触式探测技术,而实现实时数据处理、低成本、低能耗的智能电磁感知系统一直是微波探测领域的长期追求目标.为了实现这一目标,文中将能够任意调控电磁波前的可编程超表面与具有强大信息处理能力的深度学习技术相结合,实现了智能电磁感知系统;并分别从理论、系统设计与实验三个角度深入研究了深度学习驱动的智能电磁感知方法.文中首先利用“k空间”法分析了基于可编程超表面的电磁感知系统的理论分辨率和影响因素;然后介绍了一款32×24的1比特可编程超表面的系统设计,并实验验证了其动态调控电磁波前的良好性能;在此基础上将深度学习技术引入感知系统中,建立了自适应处理高维感知数据的成像卷积神经网络,实现了高保真度的人体姿态成像.本文的研究成果可用于指导智能电磁感知系统的设计与分析,并且为未来人工智能时代的电磁感知系统开辟了一条新的途径.Electromagnetic(EM)sensing has become an urgent non-contact detection technology in modern society,the realization of real-time data processing,low-cost,low-energy intelligent electromagnetic sensing system has been the long-term goal in the field of microwave detection.In order to achieve this goal,this paper combines programmable metasurface,which can arbitrarily control electromagnetic wavefront,with deep learning technology,which has powerful information processing capability,to build intelligent electromagnetic sensing system.This paper studies the deep learning driven intelligent electromagnetic sensing strategy from the aspects of theory,system design and experiment.In this paper,with the aid of“k-space”analysis,we first discuss the theoretical resolution and influencing factors of electromagnetic sensing system based on programmable metasurface.Then,the system design of a 32×241-bit programmable metasurface is introduced,and its performance of dynamic control of electromagnetic wavefront is verified by experiments.On this basis,we introduce deep learning technology into the EM sensing system,and build an imaging convolution neural network for adaptive processing of high-dimensional sensing data.Finally,the intelligent EM sensing system have achieved human posture imaging with high fidelity.These research results can be used to guide the design and analysis of intelligent electromagnetic sensing system,and open up a new way for electromagnetic sensing system in the future artificial intelligence era.

关 键 词：可编程超材料 人工智能 电磁感知 机器学习 深度学习 

分 类 号：TN802[电子电信—信息与通信工程]