An Efficient Vector-valued Belief Propagation Decoder for Quadrature Spatial Modulation

By spatially modulating the in-phase and quadrature (IQ)-components of the transmit symbols separately, quadrature spatial modulation (QSM) schemes are known to achieve almost double the spectral efficiency of conventional spatial modulation (SM) schemes. However, the much larger signal search space of QSM makes probabilistic decoding impractical, especially for popular massive multiple-input multiple-output (mMIMO) systems. In order to circumvent this challenge, we propose a novel reduced-complexity QSM decoding algorithm which utilizes the Gaussian belief propagation (GaBP) framework with two vector-valued variable nodes enabled by a sparse and IQ-decoupled reformulation of the signal model. The proposed method is shown to operate on a significantly reduced search space, while approaching the brute-force maximum likelihood (ML) estimation performance as verified by simulation results.