Symmetry-Aided Clustering Blind Detection in Spatial Modulation Systems

This paper considers the blind detection of spatial modulation multiple-input multiple-output (SM-MIMO) systems based on clustering algorithms. Firstly, we propose a novel framework (framework 1) that can guarantee the excellent detection performance of clustering detectors to implement the signal detection even if in the case of short block length by considering the rotational symmetry of digital modulation constellations. Then, we propose a novel rule by defining the concept of pseudo rotational symmetry of cluster centroids, which can check the quality of the cluster centroids of clustering algorithms. Besides, based on the proposed rule, we investigate two scenarios and propose framework 2 and framework 3 for the problems in the scenarios. Specifically, the framework 2 can early terminate the clustering algorithms that are performed multiple times different initializations to avoid poor detection performance, and it significantly reduces computational complexity while maintaining excellent performance. The framework 3 can adjust the data length of clustering algorithms according to different environmental conditions that can improve the detection efficiency while improving performance. Moreover, the proposed framework 1 can be combined with proposed framework 2 or framework 3, which further enhances the detection performance. In addition, we conduct the performance analysis for the SM-MIMO systems when the perfect channel state information (CSI) is unavailable. Simulation results demonstrate the clustering detectors combined with proposed frameworks behave better than the clustering detectors without the proposed frameworks in terms of performance or complexity.

Automated summary

This paper proposes a blind detection method for SM-MIMO systems based on clustering algorithms. By introducing novel frameworks and rules, excellent detection performance is guaranteed, and different scenarios are addressed with tailored solutions. Experimental results show that clustering detectors combined with the proposed frameworks outperform those without in terms of performance and complexity.