Massive MIMO systems, which employ large antenna arrays to serve multiple users simultaneously, generate enormous volumes of channel state information that must be acquired, processed, and fed back within strict latency constraints. Compressive sensing techniques exploit the inherent sparsity of wireless channels in appropriate representation domains to reduce the acquisition and feedback overhead. This paper presents a comprehensive comparative study of compressive sensing techniques for sparse signal recovery in massive MIMO systems. We evaluate matching pursuit algorithms, basis pursuit methods, approximate message passing, and deep learning-based recovery approaches across various channel sparsity levels, measurement ratios, and noise conditions. Experimental results demonstrate that deep learning-based recovery methods achieve 12-18% better normalized mean square error performance at low measurement ratios, while iterative algorithms provide better performance at high measurement ratios with lower computational complexity. We propose an adaptive recovery framework that selects the optimal recovery algorithm based on estimated channel sparsity and available computational resources.
Mohamed et al. (Tue,) studied this question.