LR-aided sphere decoding of generalized spatial modulation signals

Abstract

This paper presents a new low complexity sphere decoding (SD) for the generalized spatial modulation (GSM). We introduce a pre-processing stage using the lattice reduction (LR) aided minimum mean squared error (MMSE) equalization in the GSM systems. This stage speeds up the search in the decoding tree and provides a lattice dependent (LD) initial choice of the radius. Moreover, we derive a lattice independent (LI) initial radius that guarantees the optimal performance at a high signal-to-noise ratio (SNR). We also propose an iterative method to increase the radius in order to achieve the maximum likelihood (ML) performance at all SNRs. We show that the proposed algorithm achieves the ML performance while requiring prominently less computational complexity (CC) than an exhaustive search. In addition, we analyze the CC of the resulting algorithm at high SNRs and we derive an analytical expression for the complexity. The simulation results demonstrate a noticeable decrease in the CC of the proposed sphere decoding algorithm in comparison with its counterparts, particularly at low SNRs.