Spectrum Sensing Based on Sticky Hidden Markov Model With Beta Prior of Channel State Switches

Abstract

Spectrum sensing is a key functionality in cognitive radio to avoid unacceptable interference to primary users. When channel traffic becomes heavy, channel state switches within a sensing duration with a high probability, and conventional spectrum sensing methods suffer from a performance floor in miss-detection probability. In this work, after formulating a binary hypothesis problem, we develop a sticky hidden Markov model with a Beta distribution prior (SHMM-Beta) algorithm to categorize the received samples into two clusters corresponding to channel states ON and OFF. Then, we design an SHMM-Beta based detector for spectrum sensing whose test statistic is the ratio of the posterior probabilities of the hidden state variable at the last time instant of the sensing duration. Simulation results show that the proposed SHMM-Beta based detector does not exhibit performance floor in miss-detection probability, and significantly outperforms the conventional detectors specifically designed to handle the switches of channel states.