Threshold mismatch of energy detection (ED) under noise power uncertainty (NPU) model was investigated. A threshold optimization algorithm of ED (TO-ED) was proposed based on constant false alarm ratio (CFAR) criterion. Firstly, for optimizing the threshold of robust statistics approach (RSA) under the condition of CFAR, the unique existence of optimal modified factor was proved. Secondly, since the average false alarm probability of RSA is a monotonic function of modified factor, Newton dichotomy was applied to quick search of this optimal factor. Simulations showed that the performance of TO-ED is consistent with theoretical analysis and its robust detection is realized with a lower signal-to-noise ratio(SNR) comparing with RSA.
