Robust estimation of cyclic correlation in contaminated Gaussian noise

This paper considers the effect of non-Gaussian noise on the conventional estimate of cyclic correlation. It is shown that noise having a distribution function with heavier tails than the Gaussian slows the convergence of the estimate to the expected value. Alternative estimators are proposed based on the statistical concepts of robustness. These alternative estimators are shown via Monte Carlo simulation to perform well in both Gaussian and non-Gaussian noise. Another contribution of this paper is the generalization of some robust estimators to complex (versus real) data.