On the utility of sixth-order cyclic cumulants for RF signal classification

Several researchers have recently proposed the use of higher-order cumulants as features for automatic communication-signal classification. Cumulants can be associated with a stationary signal model or with a cyclostationary signal model, in which case they are periodically time-varying with Fourier components called cyclic cumulants. These researchers limit the maximum order to four, and it is commonly believed that sixth- and higher-order cumulants provide no additional benefit due to their relatively large measurement-error variance. However, we have observed experimentally that classification performance can be improved by the use of sixth-order cyclic cumulants. In this paper, we present the results of a simulation study of the cumulant variance for signals corrupted by cochannel interference and AWGN. Situations in which the use of sixth-order cumulants provides a classification benefit are provided and the benefit is quantified.