Using Hidden Markov Models to enable performance awareness and noise variance estimation for energy detection in cognitive radio

Abstract

Cognitive Radio (CR) systems are expected to enable dynamic access to the frequency spectrum in future generations of wireless communication systems by allowing unlicensed secondary users (SU) to access licensed spectrum that is not actively used by licensed primary users (PU). Due to its simplicity, energy detection is a main approach to identify spectrum opportunities in CR systems. However, the performance of energy-based spectrum sensing depends on accurate estimation of the noise energy which is required to determine the optimal sensing threshold that implies desired values for the probabilities of detection and false alarm. In this paper we study the use of Hidden Markov Models (HMM) to describe the output of energy detectors used for spectrum sensing in CR systems, and we present a novel approach that enables performance awareness and noise variance estimation for the energy detector. The proposed approach uses the values of the probabilities of detection and false alarm estimated from the HMM output to determine when the detector performance degrades due to changes in noise variance and to estimate the new variance value. Numerical results obtained from simulations are presented to illustrate application of the proposed approach.