Common Sparsity Based Channel Estimation for FDD Massive MIMO-OFDM Systems via Multitask Bayesian Compressive Sensing

In this paper, we propose a new channel estimation scheme for downlink channels in frequency division duplex (FDD) massive multiple-input multiple-output (MIMO) systems where orthogonal frequency division multiplexing (OFDM) is adopted. In this scenario, the channel sparsity level is assumed unknown to the base station (BS). To the best of our knowledge, multitask bayesian compressive sensing (MBCS) has not been used in channel estimation of FDD massive MIMO systems. By exploiting the spatially common sparsity within the system bandwidth, the MBCS can learn the common sparsity characteristics of the user channels, which guarantees the performance of sparse channel recovery. Based on MBCS, we propose a pilot adapted MBCS (PAMBCS) scheme to further exploit the sparsity feature, where the pilot sequences are designed by minimizing the differential entropy of estimated channel vectors to reduce the estimation uncertainties. Simulation results have shown that the MBCS has a good capability to reduce pilot overhead, even though when a few number of subcarriers can be used for pilot transmission. Moreover, the performance of PAMBCS is much better than random pilots based MBCS.