Deep Learning DOA Estimation Based on Transmission-Type Metasurface for Full-Amplitude Detection Array

In recent years, with the development and large-scale application of radar technology, higher requirements have been placed on the cost and accuracy of signal angle estimation. Direction of arrival (DOA) estimation is widely used in radar, sonar, communications, astronomy, and electronic reconnaissance. Against the background of the development and application of 5G/6G technologies, the demand for array signal processing is also increasing day by day. In general, high-resolution DOA estimation requires a large array aperture and multiple radio frequency chains, resulting in high system hardware cost and high power consumption. In this paper, an amplitude detection sensor array based on a transmission metasurface is designed, which uses only the amplitude information of the signal and employs a logarithmic detector for sampling. Considering the sampling error of the logarithmic detector, a deep learning algorithm is adopted. Under low signal-to-noise ratio and few snapshot conditions, the target resolution probability can reach over 95%, with a minimum resolution of 2 degrees.