A Comparative Study of Compressive Sensing Techniques for Sparse Signal Recovery in Massive MIMO

Massive multiple-input multiple-output (MIMO) systems are essential for next-generation wireless networks, but their high-dimensional signal processing demands pose challenges, particularly in sparse signal recovery. This study provides a comprehensive comparative analysis of compressive sensing (CS) techniques—optimization-based, greedy, Bayesian, learning-based, and hybrid methods—for sparse signal recovery in massive MIMO systems. The novelty lies in offering the first unified benchmarking framework under realistic conditions, including varying signal-to-noise ratios (SNR), sparsity levels, and signal dimensions. Using synthetic and real-world datasets (e.g., COST 2100), the study evaluates recovery accuracy (normalized mean squared error, NMSE), computational efficiency (runtime), robustness to noise, and scalability. Results reveal that hybrid methods achieve the best trade-off between accuracy, runtime, and noise resilience, with NMSE as low as 0.018 at 20 dB SNR and strong scalability for large signal dimensions. Learning-based methods excel in runtime performance, making them suitable for real-time applications, while Bayesian methods provide superior noise robustness. In contrast, optimization-based and greedy methods, though widely used, face computational inefficiencies and noise sensitivity in high-dimensional scenarios. These findings advance the understanding of CS techniques for massive MIMO, offering actionable insights for robust, scalable signal recovery in 5G and beyond.