Impact evaluation of noise uncertainty in spectrum sensing under middleton class a noise

Reliable spectrum sensing is required to enable the effective use of Cognitive Radio (CR) networks. In practice, CRs not only have to cope with impulsive noises but also need to consider noise uncertainties. Small uncertainties of noise parameters are inevitable in any practical detector and the impact of these must be considered. This paper considers the energy detector based spectrum sensing under Middleton Class A noise with parameter uncertainties. Firstly, we show how analytical expressions of the probabilities of detection and false alarm can be derived. Secondly, we show that the mismatch of impulsive parameters A and Γ have little impact on performance and can thus generally be ignored but the uncertainty of noise power σZ2 can induce SNR Walls to the detector, which means that, no matter how many observations are obtained, the detector cannot robustly detect the primary signal. These results have implications for designing an enhanced energy detector for robust spectrum sensing.