Pruning the Pilots: Deep Learning-Based Pilot Design and Channel Estimation for MIMO-OFDM Systems
share
With the large number of antennas and subcarriers, the overhead due to pilot transmission for channel estimation can be prohibitive in wideband massive multiple-input multiple-output (MIMO) systems. This can degrade the overall spectral efficiency significantly, and curtail the potential benefits of massive MIMO. In this paper, we propose a deep neural network (DNN)-based scheme, consisting of convolutional and dense layers, for joint pilot design and downlink channel estimation for frequency division duplex (FDD) MIMO orthogonal frequency division duplex (OFDM) systems. We also propose an effective pilot reduction technique by gradually pruning less significant neurons from the dense neural network (NN) layers during training. Our proposed NN-based scheme exploits the frequency-specific features of the propagation environment to design more efficient pilots. It outperforms linear minimum mean square error (LMMSE) estimation by learning the channel statistics over the training data and exploiting it to estimate the channel without prior assumptions on the channel distribution. Finally, our pruning-based pilot reduction technique effectively reduces the overhead by allocating pilots across subcarriers non-uniformly; allowing less pilot transmissions on subcarriers that can be satisfactorily reconstructed by the subsequent convolutional layers utilizing inter-frequency correlations.
READ FULL TEXT