Deep Learning for Compressed Sensing Based Channel Estimation in Millimeter Wave Massive MIMO

Channel estimation is considered for multi-user millimeter wave (mmWave) massive multi-input multi-output system. A deep learning compressed sensing (DLCS) channel estimation scheme is proposed, and it consists of beamspace channel amplitude estimation and channel reconstruction. The neural network (NN) for the DLCS scheme is trained offline using simulated environments according to the mmWave channel model. Then the correlation between the received signal vectors and the measurement matrix is input into the trained NN to predict the beamspace channel amplitude. Afterwards, the channel is reconstructed based on the obtained indices of dominant beamspace channel entries. Simulation results demonstrate that the proposed DLCS channel estimation scheme outperforms the existing schemes including the orthogonal matching pursuit and the distributed grid matching pursuit in terms of the normalized mean-squared error and the spectral efficiency.