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

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 32×24 1-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.