Bispectrum based estimation of parameters of Middleton Class A noise

Abstract

The proposed approach is applicable if the non-Gaussian component of the process has a non-zero mean and a skewed distribution. If the Gaussian background noise is white, it is shown that a linear relationship exists between bispectral magnitude on one frequency axis and the power spectral density of the process. Least squares estimates of the unknowns of the equation can then be obtained, from which the parameters of the process follow. Expressions for bispectrum of a stationary Poisson impulse process in terms of the rate of the process are derived. Asymptotic bias in the estimates of bispectrum of non-Gaussian processes in an additive Gaussian process is derived.<>