Detection and classification of cyclostationary signals in noise of unknown distribution is addressed and novel tests for cyclostationarity are proposed. Both cases of known and unknown signal statistics are considered. The proposed approaches exploit the asymptotic normality of sample cyclic- cumulant and polyspectrum estimators for deriving asymptotically optimal X² tests. Simpler, but generally suboptimal versions are also presented. Simulations are performed to test the proposed algorithms and illustrate their insensitivity to any stationary noise as well as the ability of higher-than second-order schemes to suppress cyclostationary Gaussian interferences of unknown covariance.

Date Published: 30 November 1992
PDF: 12 pages
Proc. SPIE 1770, Advanced Signal Processing Algorithms, Architectures, and Implementations III, (30 November 1992); doi: 10.1117/12.130939

Published in SPIE Proceedings Vol. 1770:
Advanced Signal Processing Algorithms, Architectures, and Implementations III
Franklin T. Luk, Editor(s)