Deep Learning-Based Downlink Channel Estimation for FDD Massive MIMO Systems

This letter is concerned with the downlink channel estimation in frequency division duplex (FDD) massive multiple-input multiple-output (MIMO) system. With the number of antennas increased, acquiring the downlink channel state information (CSI) becomes complex, thus restricts the performance of communication systems. A deep learning based algorithm is used to estimate the downlink CSI without the feedback. In the proposed method, the uplink cluster is firstly obtained from the receiving signals. Based on the uplink cluster data, the downlink channel is then estimated by a neural network. Simulation results show that the proposed algorithm can achieve higher achievable spectral efficiency.