STATISTICAL TESTS FOR PRESENCE OF CYCLOSTATIONARITY

Presence of kth-order cyclostationarity is defined in terms of nonvanishing cyclic-cumulants or polyspectra. Relying upon the asymptotic normality and consistency of kth-order cyclic statistics, asymptotically optimal chi2 tests are developed to detect presence of cycles in the kth-order cyclic cumulants or polyspectra, without assuming any specific distribution on the data. Constant false alarm rate tests are derived in both time- and frequency-domain and yield consistent estimates of possible cycles present in the kth-order cyclic statistics. Explicit algorithms for k less-than-or-equal-to 4 are discussed. Existing approaches are rather empirical and deal only with k less-than-or-equal-to 2 case. Simulation results are presented to confirm the performance of the given tests.