Channel quality prediction based on Bayesian inference in cognitive radio networks

Abstract

The problem of channel quality prediction in cognitive radio networks is investigated in this paper. First, the spectrum sensing process is modeled as a Non-Stationary Hidden Markov Model (NSHMM), which captures the fact that the channel state transition probability is a function of the time interval the primary user has stayed in the current state. Then the model parameters, which carry the information about the expected duration of the channel states and the spectrum sensing accuracy (detection accuracy and false alarm probability) of the SU, are estimated via Bayesian inference with Gibbs sampling. Finally, the estimated NSHMM parameters are employed to design a channel quality metric according to the predicted channel idle duration and spectrum sensing accuracy. Extensive simulation study has been performed to investigate the effectiveness of our design. The results indicate that channel ranking based on the proposed channel quality prediction mechanism captures the idle state duration of the channel and the spectrum sensing accuracy of the SUs, and provides more high quality transmission opportunities and higher successful transmission rates at shorter spectrum waiting times for dynamic spectrum access.