Gaussian Approximation Based Low Complexity Soft Demapper for Spatial Modulation Aided MIMO-LDM

Recently, spatial modulation (SM) has been integrated into multiple-input multiple-output (MIMO) layered division multiplexing (LDM) systems to further improve spectral efficiency. In this paper, we consider an SM-aided MIMO-LDM system with two layers, i.e., the core layer (CL) and the enhanced layer (EL), and propose two complexity-reduced soft demappers based on full/partial Gaussian approximation (FGA/PGA) to alleviate the computation burden of the CL demapping. The principle is to fully or partially approximate the constellation domain symbol of the low power EL signal as Gaussian inputs. By investigating the probability distribution of the equivalent noise, the FGA and PGA based demappers are proposed, respectively. It is observed that, unlike GA-based CL demapping in conventional single-input single-output (SISO) LDM systems, the received equivalent noise vector is spatially correlated due to the inherent nature of the SM signal. Simulation results are provided to validate the effectiveness of the proposed FGA and PGA based demappers. It is shown that negligibly degraded performance compared to the optimal demapper can be obtained with a greatly reduced complexity when the injection level (IL) is high or the code rate is low, and the PGA-based demapper achieves better performance than the FGA-based demapper, especially for the case of low IL and high code rate.