Hybrid Message Passing Approach for Uplink Massive MIMO Channel Estimation

Abstract

Massive MIMO channel has a sparse representation in the angular domain, but it is difficult to utilize such sparsity for uplink channel estimation due to the coupling effect caused by arbitrary pilots. The existing uplink channel estimation methods usually adopt a prior LS estimator to decouple pilots, and then recover the sparse channel with the decoupled signal. However, the prior LS estimator cannot work well for any non-orthogonal pilots, and also hinders enhancing the estimation performance by jointly decoupling pilots and exploiting the channel sparsity. To overcome the above shortcomings, we present a hybrid message passing (HMP) approach for uplink channel estimation. The novelty of HMP is to adopt the expectation propagation approximation (EPA) to decouple pilots instead of using the prior LS estimator, and then utilize the generalized approximate message passing (GAMP) to recover the sparse channel. Since the two procedures work iteratively, HMP can jointly decouple pilots and exploit the channel sparsity. Numerical simulation results verify the superiority of HMP.