A Semi-Passive RIS Channel Estimation Method based on Super-Resolution Network

Reconfigurable intelligent surface(RIS) has become a novel enabling technology of 6G. It is widely used in coverage enhancement, signal regulation and other communication scenarios due to its low cost and low energy consumption. This paper considers the RIS-assisted millimeter-wave communication system, deploys some active elements on the RIS, and proposes a channel state information (CSI) reconstruction scheme based on fast super-resolution convolutional neural network (FSRCNN). Sub-pixel layer is used to replace deconvolution layer, and low-resolution images are directly input to the network. In addition, the noise of the CSI is further reduced by denoising convolutional neural network(DnCNN). Experiments show that the proposed scheme can recover CSI well.