Title: Deep Learning Based FDD Non-Stationary Massive MIMO Downlink Channel Reconstruction

Abstract: This paper proposes a model-driven deep learning-based downlink channel reconstruction scheme for frequency division duplexing (FDD) massive multi-input multi-output (MIMO) systems. The spatial non-stationarity, which is the key feature of the future extremely large aperture massive MIMO system, is considered. Instead of the channel matrix, the channel model parameters are learned by neural networks to save the overhead and improve the accuracy of channel reconstruction. By viewing the channel as an image, we introduce You Only Look Once (YOLO), a powerful neural network for object detection, to enable a rapid estimation process of the model parameters, including the detection of angles and delays of the paths and the identification of visibility regions of the scatterers. The deep learning-based scheme avoids the complicated iterative process introduced by the algorithm-based parameter extraction methods. A low-complexity algorithm-based refiner further refines the YOLO estimates toward high accuracy. Given the efficiency of model-driven deep learning and the combination of neural network and algorithm, the proposed scheme can rapidly and accurately reconstruct the non-stationary downlink channel. Moreover, the proposed scheme is also applicable to widely concerned stationary systems and achieves comparable reconstruction accuracy as an algorithm-based method with greatly reduced time consumption.