Cognitive Radio Spectrum Detection Based On Bayes Inference

With the rapid development of wireless communication, spectrum resource has become scarce. Cognitive Radio (CR) is the very technology to solve this problem, which exploits spectrum resource efficiently. As for CR, spectrum detection is the very technology upon which the entire operation of cognitive radio rests. The performance of general spectrum detection algorithms which detect a certain band in one instant is affected by shadow effect and so on. The probability of false detection is lowered by observing the band consecutively and estimating the primary user’s states. In this paper, the technology of spectrum detection in CR is first introduced briefly, and Bayes spectrum detection methods are researched afterwards. In the end, a method of state estimation for a primary user based on Variational Bayes (VB) is proposed. Because the primary user’s states are hidden and can only be estimated by observation, the system is modeled as a hidden Markov model (HMM) with Dirichlet priors. After that, VB algorithm is applied in the evaluation of a primary user’s states to estimate the model’s parameters, which solves the intractability of Bayes’rule when estimating the parameters in HMM and avoids probable overfitting problem. The primary user’s states are obtained by using Viterbi algorithm. Simulation shows that the proposed state estimation algorithm for a primary user based on VB in this paper outperforms.