On the Accuracy of Maximum Likelihood Estimation for Primary User Behavior in Cognitive Radio Networks

The primary user (PU)'s busy/idle behavior in a cognitive radio network is conventionally modeled using a two-state Markov chain. Maximum likelihood (ML) estimation is widely applied to estimate the state transition probabilities. This letter derives a precise expression of the probability mass function (PMF) for the ML estimator, which has not been reported in the literature. By leveraging the exact PMF expression, the essential relation among the number of samples, transition probabilities, and estimation accuracy is revealed.