Autocorrelation-Based Traffic Pattern Classification for Cognitive Radios

This paper proposes a autocorrelation-based method to classify traffic patterns of primary channels in cognitive radio systems to allow a more accurate prediction of the future idle times. The classification algorithm uses binary information collected by spectrum sensing. It searches periodicity from the sensed binary pattern using a discrete autocorrelation function. Errors that are caused by noise and possible false sensing reports are filtered away from the autocorrelation function. We tested the method with Pare to, Weibull, and exponentially distributed stochastic traffic, and with deterministic traffic. The proposed method finds the type of traffic with a high probability when the channels of interest include both stochastic and deterministic traffic. Stochastic traffic is always classified right and regarding the deterministic traffic the probability of correct classification is over 95% when the probability of missed detection or probability of false alarms is below 10%.