eduzhai > Applied Sciences > Engineering >

Pruning the Pilots Deep Learning-Based Pilot Design and Channel Estimation for MIMO-OFDM Systems

  • Save

... pages left unread,continue reading

Document pages: 28 pages

Abstract: With the large number of antennas and subcarriers the overhead due to pilottransmission for channel estimation can be prohibitive in wideband massivemultiple-input multiple-output (MIMO) systems. This can degrade the overallspectral efficiency significantly, and as a result, curtail the potentialbenefits of massive MIMO. In this paper, we propose a neural network (NN)-basedjoint pilot design and downlink channel estimation scheme for frequencydivision duplex (FDD) MIMO orthogonal frequency division multiplex (OFDM)systems. The proposed NN architecture uses fully connected layers forfrequency-aware pilot design, and outperforms linear minimum mean square error(LMMSE) estimation by exploiting inherent correlations in MIMO channel matricesutilizing convolutional NN layers. Our proposed NN architecture uses anon-local attention module to learn longer range correlations in the channelmatrix to further improve the channel estimation performance. We also proposean effective pilot reduction technique by gradually pruning less significantneurons from the dense NN layers during training. This constitutes a novelapplication of NN pruning to reduce the pilot transmission overhead. Ourpruning-based pilot reduction technique reduces the overhead by allocatingpilots across subcarriers non-uniformly and exploiting the inter-frequency andinter-antenna correlations in the channel matrix efficiently throughconvolutional layers and attention module.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...