State estimation for a primary user in Cognitive Radio based on Variational Bayes

Variational Bayes (VB) is applied in Cognitive Radio (CR) to handle the intractability of Bayes´ rule in Hidden Markov Model (HMM) when we evaluate parameters to estimate a primary user´s (PU´s) states but without any information on a PU. It can also avoid a probable overfitting problem caused by maximum likelihood (ML) algorithm. With the advantage of conjugating to complete-data likelihood in HMM of CR, Dirichlet priors are chosen as priors in HMM for VB. By using Viterbi algorithm to decode, the performance of the proposed algorithm is evaluated by simulation with similar performance by using the known parameters.