Real-Time Adaptively Regularized Compressive Sensing in Cognitive Radio Networks

Abstract

Wideband spectrum sensing is regarded as one of the key functional blocks in cognitive radio systems, where compressive sensing (CS) has become one of the promising techniques to deal with the Nyquist sampling rate bottleneck. Theoretical analyses and simulations have shown that CS could achieve both high detection and low false alarm probabilities in wideband spectrum sensing. However, the implementation of CS over real-world signals and real-time processing poses significant challenges due to the high computational burden and reconstruction errors against noise. In this paper, we propose an efficient adaptively regularized iterative reweighted least squares (AR-IRLS) algorithm to implement the real-time signal recovery in CS-based wideband spectrum sensing. The proposed AR-IRLS algorithm moves the estimated solutions along an exponential-linear path by regularizing weights with a series of nonincreasing penalty terms, which significantly speeds up the convergence of reconstruction and provides a high fidelity guarantee to cope with spectral signals with varying bandwidths and power levels. Furthermore, a descent-based decision threshold setting algorithm is proposed to distinguish the primary signals from the mixture of the reconstruction errors and unknown noises. The proposed scheme demonstrates robustness against different sparsity levels at low compressive ratios without degradation of the reconstruction performance. It is tested with the real-world signals over the TV white space after being validated with the simulated signals. Both the simulation and real-time experiments show that the proposed scheme outperforms the conventional iterative reweighted least squares algorithms in terms of convergence speed, reconstruction accuracy, and compressive ratio.