Spectrum Prediction in Cognitive Radio Networks: A Bayesian Approach

Spectrum sensing is an important activity in the Cognitive Radio (CR) scenario. Presence of primary users (PU) over a specific band has to be monitored periodically in each time slots. Spectrum sensing and data communication have to be completed in each time slot. If we can reduce the time required for spectrum sensing, more data can be transmitted in the specified time slot. If the presence of a PU can be predicted by a CR, throughput of the system can be improved. In this paper we define a simple approach based on Bayesian theorem to predict spectrum occupancy status of PU, from its spectrum occupancy pattern. Its performance is compared with exponential weighted moving average (EWMA) based approach to predict the spectrum occupancy information. A modification to EWMA is also suggested named hybrid approach by including the Bayesian probability within the above approach. Their performance is compared in terms of predicted probability and spectrum decision. Spectrum decision at different parameters of beta distribution is compared. Impact of number of previous data considered for prediction is also explored. Bit error rate of Bayesian approach is found less at certain data distributions. Computational requirement of Bayesian approach is also found relatively less.