Deep Learning Assisted Channel Estimation Refinement in Uplink OFDM Systems Under Time-Varying Channels**This work was supported in part by the National Natural Science Foundation of China (No. 61871327, 61801218 and 61701407), the Natural Science Basic Research Plan in Shaanxi Province of China (No.2018JM6037 and 2018JQ6017),

In various practical orthogonal frequency-division multiplexing (OFDM) systems, the estimation accuracy at the receiver is challenging, and, specifically when operate over time-varying channels. This occurs mostly due to the presence of multipath Doppler shifts. Meanwhile, deep learning has quite recently demonstrated its superiority in extracting features information from big data. To this end, in this paper, a deep learning-assisted approach for channel estimation refinement is proposed in OFDM systems, under uplink time-varying channels. By exploitingfully-connected deep neural network (FC-DNN) properly, we successfully design a channel parameter refine network (CPR-Net) which combines deep learning with existing channel estimation algorithms. Simulation results demonstrate that, compared with conventional channel estimation algorithms, the proposed CPR-Net can significantly improve the estimation accuracy of channel parameters and provide more accurate and robust signal recovery performance.