Gaussian noise blind power spectrum estimation from higher order spectra
Erdal Turkbeyler, Anthony George Constantinides · 2002
Signal measurements are generally corrupted by some form of noise which is normally taken to be additive noise. Such noise degrades conventional power spectrum estimates. Higher order statistics, however offer a method for power spectrum estimation for which the effect of additive Gaussian noise is eliminated. A new method based on higher order statistics is proposed to estimate the power spectrum. The method employs the trispectrum and bispectrum to calculate the power spectrum, and correspondingly the autocorrelations. Nonparametric and parametric methods (AR, MA, ARMA models) can be employed to estimate the trispectrum and bispectrum, but in this paper, nonparametric bispectrum and trispectrum estimation methods are used. Simulation studies are presented which compare the method with conventional techniques.>