Mobility-Aware Reconstruction Algorithm for Correlated Nonzero Neighborhood Structured Downlink Channel in Massive MIMO

This paper investigates downlink (DL) channel estimation of frequency division duplex (FDD) massive multiple-input-multiple-output (MIMO) with a finite number of base station (BS) antennas. Based on the performance analysis of DL channel in FDD, it is demonstrated that the interference effect cannot be ignored in the finite region. Also, excessive training overhead is required in conventional channel estimation methods when the number of BS antennas goes large. Therefore, it is necessary to design an algorithm for alleviating the interference in addition to the reduction of pilot overhead. In this paper, a general model is presented to exploit certain sparsity structure of massive MIMO channel in beam domain, which is called nonzero-neighborhood (NZN) structure. Considering the smooth variation of channel paths during two consecutive estimation intervals, a compressed sensing-based algorithm is proposed to efficiently estimate and track NZN-structured DL channel. The convergence behavior and computational complexity of the proposed mobility-aware subspace pursuit (MA-SP) algorithm are analyzed based on restricted isometric property. It is demonstrated that the application of proposed MA-SP algorithm efficiently reduces the required pilot length in channel estimation. Interference analysis shows that inter-cell interference is considerably alleviated in the proposed channel estimation phase without any prior information about interference. Simulation results demonstrate the effectiveness of the proposed MA-SP technique compared with the previous methods.