Continuous hidden Markov model based interference-aware cognitive radio spectrum occupancy prediction

In this paper, a novel time-frequency spectrum occupancy prediction scheme in a cognitive radio network is presented which is based on a continuous hidden Markov model (CHMM) as opposed to the conventional discrete hidden Markov model (DHMM) based approaches. The primary user signal observations at a secondary user receiver equipped with an analysis filter bank for spectrum sensing are assumed to be embedded in non-Gaussian co-channel interference. The prediction scheme is carried out in three phases comprising a pilot phase, a learning phase and a prediction phase. The performance of the CHMM based scheme is investigated in different occupancy traffic scenarios and it is shown that for a sufficient length of the learning phase, the prediction can overcome the performance limitation of its DHMM counterpart resulting from quantization of observations at the receiver.