High SNR Performance Analysis of F-ESPRIT

Paper in proceeding, 2004

A performance analysis of a recently derived method for harmonic retrieval is conducted. The algorithm, which is referred to as F-ESPRIT, has a frequency selective feature which enables the user to select data from frequency 4 sub-bands in which the signal harmonics are known to reside This way an efficient suppression of out-of-band disturbances is obtained without the need of explicit data pre-filtering. A first order perturbation analysis is conducted, which results in a closed 3 form expression of the mean square error of the parameter estimates. The derived expression is validated empirically by Monte Carlo simulations and the Cramer-Rao lower bound.