Deep Learning-Based Massive MIMO CSI Feedback

Massive multi-input and multi-output technology is a key technology for future 5G wireless communication. The channel feedback problem of massive mimo becomes more and more challenging as the size of the mimo channel matrix becomes larger. A supervised deep learning-based encoder-decoder scheme was proposed to improve recinstruction quality recovery channel state information.Compared with the traditional compression-based sensing algorithm, Residual Attention-Net can still maintain good performance when compression is low.