Low-Complexity Joint Weighted Neumann Series and Gauss–Seidel Soft-Output Detection for Massive MIMO Systems

We introduce a joint weighted Neumann series (WNS) and Gauss–Seidel (GS) approach to implement an approximated linear minimum mean-squared error (LMMSE) detector for uplink massive multiple-input multiple-output (M-MIMO) systems. first propose initialize the GS iteration by WNS method, which produces closer-to-LMMSE initial solution than conventional zero vector diagonal-matrix based scheme. Then algorithm is applied LMMSE detection iteratively. Furthermore, on WNS, we devise low-complexity approximate log-likelihood ratios (LLRs) computation method whose performance loss negligible compared with exact method. Numerical results illustrate that proposed WNS-GS outperforms achieves near-LMMSE significantly lower computational complexity.