Wideband Spectrum Sensing for Cognitive Radios in Weakly Correlated Non-Gaussian Noise

Wideband signals can be subjected to various types of non-Gaussian and impulsive noise. Such non-Gaussian noise may cause degradation in the detection performance of conventional wideband signal detection methods, such as the periodogram-based approach. In addition, the signal detection probability can be further reduced due to possible correlation among the noise samples. Under such non-Gaussian noise conditions, we formulate a wideband signal detection method for cognitive radios (CRs) based on a locally-optimal (LO) Neyman-Pearson (NP) detector by assuming a weakly correlated noise model with known parameters. The corresponding decision statistic is expressed in the frequency-domain, allowing to detect the spectral activities within the sensed band of interest. The proposed detector is shown to reduce the impact of correlated non-Gaussian noise on the detection performance. We compute the receiver operating characteristic (ROC) of the proposed wideband LO-NP detector and show its superior performance, compared to existing wideband detectors under correlated noise conditions.