Impact of correlated primary transmissions on the design of a cognitive radio inference engine

We consider sensing for cognitive network users, in particular focusing on a scenario where a primary user (PU) and a secondary user (SU) operate on the same frequency band. The SU is interested in identifying transmission opportunities when the PU is silent. We investigate how this sensing performed by the SU can be improved through modeling the PU transmission pattern with increasing accuracy. In particular, we are interested in evaluating the impact of correlation in PU's transmissions. Therefore, we assume that the real behavior of the PU follows a Markov chain, used to model correlation in its activity, and we discuss how the maximum likelihood estimation of the SU can be subsequently improved by adding more information about this underlying process. In this way, the estimate can evolve into a maximum a-posteriori criterion, and furthermore knowledge about the whole Markov chain can be exploited. Also, we investigate the practical setup of training periods of variable length used to estimate the PU's parameters.