Optimal Spectrum Sensing Threshold for Unequal Priors Case

Abstract

Classical spectrum sensing techniques utilize maximum likelihood (ML) detection for identification of spectrum holes. The approach is sub-optimal for the case of un-equal priors where the probabilities of channel occupation and vacancy are not the same. Such situations are bound to occur in most commercial bands such as GSM etc and hence are of more interest. The loss in performance has been disregarded as negligible in most of the work done on spectrum sensing techniques. This paper quantifies the effects of changing priors on classical energy detection and infers that the loss in spectrum sensing performance is not negligible. The deterioration is especially considerable at low SNR values and at low probabilities of channel occupation. An optimum threshold for achieving minimum probability of error has been derived in this work for unequal prior case. Detection based on the proposed threshold out-performs classical detectors under the assumption that priors are known at the receiver.