Downlink Channel Estimation in Multiuser Massive MIMO With Hidden Markovian Sparsity

Recently, compressive sensing based massive multiple input multiple output (MIMO) channel estimation (CE) has attracted intensive research interest. By exploiting the structured sparsity of massive MIMO channels, it is possible to design structured sparse channel estimation (SSCE) algorithms to significantly reduce the pilot/feedback overheads in massive MIMO downlink CE. However, most existing SSCE algorithms are designed based on oversimplified channel models with restrictive assumptions, and thus perform poorly under realistic channels. In this paper, we propose aHidden-Markov-Model(HMM) to capture the structured sparsity of multiuser massive MIMO channels. The HMM has the flexibility to model different propagation environments that occur in practice. We derive an efficient approximate message passing algorithm called Turbo-orthogonal approximate message passing (OAMP) to solve the resulting sparse CE problem with anHMM prior. The proposed Turbo-OAMP does not require knowledge of the HMM channel parameters, which is automatically learned based on the expectation maximization framework. As a result, the proposed Turbo-OAMP algorithm derived from the HMM works well for realistic channels with complicated propagation environments and unknown channel parameters. Extensive simulations verify that the proposed Turbo-OAMP can achieve significant gains over the existing SSCE algorithms under realistic channels.