Optimal Compressive Spectrum Sensing Based on Sparsity Order Estimation in Wideband Cognitive Radios

Abstract

There are several advantages of compressive sensing (CS) technologies for spectrum sensing in wideband cognitive radios (CRs), especially in reducing the sampling rate. However, in CS, the number of measurements needed to reconstruct the sparse signal is determined by the actual sparsity order of the signal, which, in practice, is often unknown and dynamically varying. Thus, in practice, an excessive number of measurements has to be chosen, and choosing such a number degrades the sensing performance. To solve this problem, this article proposes a two-step adaptive compressive spectrum sensing (TS-ACSS) method to estimate the sparsity order. By using a simple statistical analysis of the compressive measurements, a coarse estimation of the sparsity order can be obtained, and an accurate estimation can be acquired by comparing multiple iterations. Compared with existing CS approaches, the proposed TS-ACSS method can not only adaptively adopt an optimal number of measurements without a priori knowledge about current sparsity order but can also detect a sensing failure that can happen if the wideband signal is not sparse in the frequency domain and if the noise level is too large to sense successfully. Furthermore, we verify the effectiveness of our method in dynamic spectrum changing scenarios. The results show that our method can work in both low signal-to-noise ratios (SNRs) and high SNRs environments, and can also be applied in the condition of different energy coefficients and positions of channel occupancy.