A Robust Threshold Optimization Approach for Energy Detection Based Spectrum Sensing with Noise Uncertainty

With noise power uncertainty in energy detection, fixed threshold is no longer suitable. Reviewing the conventional robust statistic approach (RSA), we show that RSA does not provide optimal threshold for primary user (PU) detection. In this paper, we develop a novel threshold optimization approach based on sensing statistical model. Adopting linear integral approximation (LIA), we present new closed form expressions for the detector's performances. Relying on constant false alarm rate (CFAR) principle and dichotomy iteration, we can obtain the optimized threshold, which is more noise-environment adaptive. Simulation results show that the detection probability can be significant improved using the optimized threshold, and the overall performance satisfies the robustness requirements.