An efficient and robust approach for wideband compressive spectrum sensing

Compressive sensing (CS), which exploits the spectral sparsity of wideband signals, is a powerful approach to wideband spectrum acquisition at sub-Nyquist sampling rates. In this paper, we propose a modified CoSaMP method for fast and accurate wideband compressive spectrum sensing in noisy environments. We exploit the priori knowledge of the noise level to enhance the robustness of spectrum reconstruction to noise. In addition, to improve the efficiency of spectrum sensing, we utilize the residual change to adjust halting conditions and introduce feasible fast solutions to the essential step least-squares estimation. Simulation results show that the modified CoSaMP algorithm outperforms the original CoSaMP and the generic convex relaxation algorithm basis pursuit denoising (BPDN) with respect to approximation performance and robustness. Moreover, it executes much faster than BPDN and CoSaMP.