Study on frequency resolution of spectral estimators using non-linear functions of the covariance matrix

Abstract

This paper studies the frequency resolution of two existing spectral estimators, where one is a generalization of the minimum variance spectral estimators (GMVSE) and the other is Pisarenko's non-linear spectral estimators (PNLSE). By this study, we show the uncovered relationship between the non-linear parameter and the frequency resolution for both the GMVSE and the PNLSE. The results bring out a clear guideline on choosing the parameters for both the GMVSE and the PNLSE to avoid missing any sinusoids.