Unsupervised Prediction of Channel State for Cognitive Radio Using Hidden Markov Model

The accurate modeling of primary users (Pus) behavior is important and crucial to cognitive radio (CR). The method to detect idle frequencies, not used by primary users' (Pus') has been widely investigated recent years. Existing researches need to estimate and select the threshold of the energy detector manually. In this paper, we propose an unsupervised approach to estimate channel states. We adopt different number of observed state according to different classification in hidden Markov model (HMM). We trained and tested the model through experiments using real spectrum measurement data. The system we proposed can automatically deal with large amounts of data and present high performance and good expansibility to predict channel state.