A Hybrid Approach for Multi-User Massive MIMO Sparse Channel Estimation Based on Bayesian Recovery and Hard Thresholding

We propose an efficient sparse channel estimation algorithm based on the compressed sensing (CS) approach for large scale multi-user (MU) MIMO systems. The proposed scheme is a hybrid one comprising Bayesian and greedy methods. It can improve the estimation performance by incorporating the spatial channel knowledge that the neighboring antennas in an array share the same support. The pilot overhead can be reduced by utilizing the data symbols using a reliability measure for channel estimation. Moreover, the effect of interfering and non-interfering pilots on the estimation performance will be investigated. It will be shown that the proposed hybrid technique performs similar or better than the Bayesian method with substantially reduced complexity.