Primary user activity prediction using the hidden Markov model in cognitive radio networks

Cognitive radio (CR) is a system for sense and access of spectrum opportunistically. It is designed on spectrum holes in primary users (PU) over licensed frequency bands. Determining access time for the secondary user (SU) is one of the most important issues in cognitive radio systems. This spectrum availability can be optimized by applying learning methods. In this paper, the hidden Markov model (HMM) is applied to determine and predict channel activity patterns. Specifically, a sensing frame structure is proposed to learn the channel activity pattern and apply the patterns as training vectors; afterward, the HMM model is modified for predicting the channel usage activity by PU. Three traffic patterns are considered as Heavy Traffic, Balanced Traffic and Slow Traffic. The results indicate 72% validity in Balanced Traffic while unbalanced traffic decreases prediction validity to 56%.