FDD massive MIMO downlink channel estimation with complex hybrid generalized approximate message passing algorithm

Abstract

Precise channel state information (CSI) is essential for the massive multiple-input multiple-output (MIMO) system to achieve high spectrum and energy efficiency performance in the forthcoming 5G communication. Combined with angular domain channel sparsity, compressive sensing (CS) technique is introduced to estimate downlink CSI because it can help the frequency division duplex massive MIMO system to overcome the restriction of the limited pilot overhead. Conventional CS techniques consider different entries of the sparse signal as equivalent random variables. However, some scatters around the base station are fixed during practical propagation. Consequently, some propagation beams are more likely to locate within certain angular spread and the corresponding entries of the angular domain channel response vector are more likely to be non-zero valued. While this non-zero probability of a certain entry can be acquired offline by learning and analyzing the historical CSI, it is unnecessary to be estimated again during the sparse reconstruction process. When we describe the non-zero probability of a certain entry with the probabilistic density manner, hybrid prior probabilistic settings are used because of the practical propagation property. Combined with the complex generalized approximated message passing (GAMP) algorithm, a new channel estimation method is introduced in this paper. We define the GAMP algorithm with its intrinsic architecture and amended hybrid probability node settings as hybrid GAMP algorithm. A definite improvement of the pilot consumption as well as the estimation accuracy are simultaneously achieved through our proposed channel estimation method with complex hybrid GAMP algorithm and accurate hybrid settings. By further simulation, the influence of the inaccurate hybrid settings of the sparse channel response vector is drawn that the negative effect is proved to be quite small for the proposed channel estimation method.