Spectrum Usage Prediction Based on High-order Markov Model for Cognitive Radio Networks

Cognitive radio is a dynamic spectrum access technology as a solution to spectrum under-utilization problem in some licensed bands. Cognitive radio should sense the spectrum usage steadily to prevent interfering licensed users and spectrum occupancy information of licensed users can be used both for learning the usage and prediction of the future occupancy. In this paper, a two-state high-order Markov chain based prediction model is presented for cognitive radio system to predict spectrum occupancy. This model adopts an improved LZ78 algorithm to evaluate transition probabilities of the two-state high-order Markov chain. Simulation results show that the proposed scheme can predict spectrum usage effectively and significantly reduce computational cost compared to prediction algorithms based on AR model.