Block Bayesian matching pursuit based channel estimation for FDD massive MIMO system

Massive multi-input multi-output (MIMO) has been recognized as a key technology for 5G communication system due to the high array gain and spectral efficiency. To redeem the merits, the base station (BS) requires the accurate channel state information (CSI), which consumes large amount of pilot overhead for downlink CSI acquisition in frequency division duplex system. To address this issue, a block Bayesian matching pursuit (BBMP) based channel estimation approach is proposed, which fully exploits the common support of channels from multiple antennas at the BS. Specifically, the channel estimation problem is formulated as the block sparse recovery problem, where the prior probability for block support set is accordingly derived. Then, based on the prior knowledge and equivalent sensing matrix, a selection metric is defined to select the block index into the dominant support by exploiting the maximum likelihood criterion. Furthermore, by utilizing a matching pursuit based approach, the update criterion is derived for selection metric when the dominant support set augments with the iteration. Simulation results demonstrate that the proposed BBMP algorithm is capable to significantly reduce the pilot overhead for channel estimation and robust to the channel sparsity information.