Compressed Sensing Based Channel Estimation in FDD Multi-User Massive MIMO Using Angle Domain Sparsity and Transmit Antenna Correlation

摘要

We study channel estimation in downlink Frequency Division Duplex multi-user massive MIMO systems. Due to the large number of antennas in massive MIMO, both channel estimation at the users, and channel state information (CSI) feedback to the base station (BS) is a challenging task. Here we use sparsity in the angle domain and the transmit antenna correlation for compressed sensing based channel estimation. We use an approach where the measurements from a group of users with similar transmit covariance matrices are fed back to BS, and BS does joint channel estimation using the multiple measurement vector (MMV) approach. We start with the physical channel model with scatterer clusters and then reduce it to the virtual channel model, where a sparse channel matrix is used, together with the angular domain unitary matrices. The users that share common clusters have common sparsity and use joint (MMV) estimation. Due to the channel structure with scatterer clusters, resulting in groups of channel coefficients, we also use group sparsity and joint group sparsity to improve estimation error. We use antenna selection to transmit training symbols at selected antennas only, which, together with the virtual channel model, results in partial DFT measurement matrix.