Robust Methods for Wideband Compressive Spectrum Sensing under Non-Gaussian Noise

Abstract

In a cognitive radio network, non-reconstruction-based wideband compressive spectrum sensing poses challenges under the non-Gaussian noise environment. The maximum correntropy criterion (MCC) is robust to impulsive noise whereas, the Parzen window Renyi entropy is a good choice for spectrum sensing in the presence of Gaussian noise at a low signal-to-noise ratio (SNR). However, the detection performance of these algorithms depends on the accuracy of measured noise variance, which is sensitive to impulse noise. In this letter, we improve the sensing performance of both the aforementioned algorithms in a non-Gaussian noise environment by modifying the kernel and threshold using robust statistics. The robust technique minimizes the influence of impulsive noise in the received signal. Finally, we carry out the simulation results to illustrate the superior performance of robust sensing algorithms under both Bernoulli's distribution and symmetric $\alpha $ stable ( $\text{S}\alpha \text{S}$ ) distribution channel noise. The performance is compared with the non-robust counterpart for sensing multi-carrier Universal-Filtered Multi-Carrier (UFMC) signal.