Bayesian Approach to Spectrum Sensing for Cognitive Radio Applications

In this paper, we address the spectrum sensing task of cognitive radio from Bayesian detection (BD) perspective. We first show that BD essentially simplifies to classical energy detection (ED) under Gaussian signal assumption but the threshold setting has more degrees of freedom to optimize the sensing performance, e.g., against given spectrum utilization. Then we propose a novel BD based algorithm where the sample energy is calculated iteratively, and the odds ratio is used to quantify the measurement reliability. Depending on the reliability, either a hard decision is forced or the algorithm progresses to accumulate more sample energy. When working under unknown SNRs, this allows the detector to reach reliable sensing decisions by using adaptive sample window, thus providing advantage over classical ED where fixed threshold is used regardless of channel conditions. Extensive computer simulations are provided to illustrate the performance advantages against classical ED in terms of e.g. sensing time.