Low-Complexity Signal Detection for Rank-Deficient GSM-MIMO

Abstract

Generalized spatial modulation (GSM) is an appealing multiple-antenna transmission paradigm, which offers a favorable trade-off between spectral efficiency and hardware complexity by activating part of the transmit antennas. However, in high-throughput scenarios characterized by rank-deficient multiple-input multiple-output (MIMO) channels, conventional linear and sub-optimal detectors suffer from pronounced performance degradation with respect to the maximum-likelihood (ML) benchmark. To address this fundamental limitation, this letter proposes an ordered-block generalized sphere decoder (OB-GSD), which combines regularization with likelihood-driven transmitantenna-combination (TAC) ordering and an early termination strategy. By judiciously reshaping and traversing the search space, the proposed OB-GSD is capable of approaching ML detection performance at a substantially reduced computational cost. Simulation results confirm that OB-GSD achieves near-ML bit-error-rate (BER) performance, while imposing a markedly lower complexity in rank-deficient GSM-MIMO systems.