An Efficient Block Sparse Recovery Algorithm for Channel Estimation

In wireless massive multiple-input multiple-output (MIMO) systems, one of the primary challenges lies in accurate and efficient acquisition of channel state information (CSI), which can be formulated as a compressive sensing (CS) problem. The current sparse Bayesian learning (SBL) framework in CS can conveniently utilize prior information in the signal, such as the block-sparse structure, to achieve superior estimation performance. However, due to matrix inversion, the computational complexity for sparse recovery is extremely high, especially in the high-dimensional applications like massive MIMO. In this paper, an efficient block sparse recovery algorithm is designed based on pattern-coupled SBL (PC-SBL), with the E-step simplified by an iterative approximate message passing (AMP) process. Simulation experiments prove that the proposed algorithm can quickly estimate the channel with high accuracy and extremely low complexity.