Classification of co-channel communication signals using cyclic cumulants

Traditional methods of signal classification, including phase and frequency histograms, modulus measurements, and power-spectrum measurements, fail when the signal-to-noise ratio is sufficiently low or when there are interfering signals present. These methods fail because the interfering signals and noise contribute substantially to the measured values of the classification features, thereby obscuring the contribution to the measurement from the signal of interest. The required signal selectivity of classification features for this situation can, in some instances be provided by features based on the cyclostationarity of both the signal of interest and the interferers. A set of cyclic-cumulant-based features for signal classification is proposed and analyzed, and results of classification experiments using simulated data are presented. The simulation results reveal that each of a number of spectrally overlapping signals can be successfully classified by measuring and processing the proposed features.