Semantic Diversity for Massive MIMO CSI Feedback

Massive multiple-input multiple-output (MIMO) technology enhances throughput by relying on accurate channel state information (CSI) at the base station (BS). However, bandwidth constraints and low signal-to-noise ratio (SNR) conditions often lead to errors that degrade CSI reconstruction. To address this, semantic communication has emerged as a promising solution, exhibiting superior performance in low-SNR environments. This paper introduces an adaptive coding frame-work that considers the varying significance of semantic symbols, allowing for the prioritization of critical symbols based on their spatial positions within the communication system. Building on this framework, a novel semantic diversity method is proposed to enhance the robustness of CSI reconstruction in challenging SNR conditions. Furthermore, we present a multi-level semantic diversity strategy that extends these benefits, optimizing system performance across a wider SNR range. Experimental results demonstrate that the proposed approach not only improves CSI reconstruction accuracy but also provides scalable performance gains across different feedback bandwidths.