Deep Learning-Based Implicit CSI Feedback in Massive MIMO

Massive multiple-input multiple-output can obtain more performance gain by exploiting the downlink channel state information (CSI) at base station (BS). Therefore, studying CSI feedback with limited communication resources in frequency-division duplexing systems is of great importance. Recently, deep learning (DL)-based has shown considerable potential. However, existing DL-based explicit schemes are difficult to deploy because current fifth-generation mobile protocols and designed based on an implicit mechanism. In this paper, we propose a architecture inherit low-overhead characteristic, which uses neural networks (NNs) replace precoding matrix indicator (PMI) encoding decoding modules. By using environment information, NNs achieve refined mapping between PMI compared codebooks. The correlation subbands also used further improve performance. Simulation results show that, for single resource block (RB), proposed save 25.0% – 40.0% overhead Type I codebook under different antenna configurations. For wideband system 52 RBs, be saved 30.7% 48.0% II when ignoring considering extracting subband correlation, respectively.