Comparative Performance Analysis of Sparse Recovery Algorithms
Over the past five years, the compressed sensing (CS) recovery has increased considerably owing to its applications in the fifth-generation (5G) and sixth-generation (6G) massive Multiple Input Multiple Output (MIMO) channel estimation. Moreover reconfigurable intelligent surface (RIS) has emerged as an essential framework for 6G wireless communication systems. This limitation results in a highly underdetermined estimation problem, exacerbated by the large dimensionality of the IRS-induced channel matrices and the necessity to estimate the cascaded base station (BS) IRS user equipment (UE) links. The main concept behind Compressed Sensing is that of recovering sparse signals from much fewer observations under some specific conditions. This idea is exploited to develop a variety of recovery algorithms applied in signal processing and wireless communications. This paper presents a systematic comparative study of three CS-based recovery algorithms: Orthogonal Matching Pursuit (OMP), Structured Matching Pursuit (StrMP), and Dictionary-based Sparse Bayesian Learning (DSBL). To evaluate the system performance, Normalized Mean Square Error (NMSE) and runtime over Monte Carlo trials are used. The simulation results show that the OMP outperformed DSBL at all SNR values. On the other hand DSBL does not offer competitive accuracy in this setting, but its automatic sparsity estimation remains an advantage when is unknown.. StrMP failed under random sparsity. Connections to deep unfolding and Bayesian generative models are discussed.
Authors
- Muamer Hawej (ORCID: https://orcid.org/0000-0002-3875-6124)
- Hunaydah A. Qadour
Institutions
- University Elmergib (LY)
Publication Details
- Journal
- Comprehensive Journal of Science
- Published
- 2026-11-15
- DOI
- https://doi.org/10.65405/nav34k90
- Primary Topic
- Sparse and Compressive Sensing Techniques
- Type
- article
- Field-Weighted Citation Impact
- 0.00