PARAMETRIC ESTIMATION OF SINUSOIDS IN NOISE A comparison between parametric approaches and the definition of a regularized Smyth algorithm

摘要

A comparison between well-established parametric algorithms and the more recent Smyth algorithm for estimation of sinusoidal signals in white noise is presented. The comparison is performed through a pseudo-Monte Carlo analysis on simulated data. The results obtained show that Smyth algorithm has a slightly better performance at large Signal-to Noise Ratios. However, when the SNR drops down, the performance of the Smyth algorithm dramatically decreases. A better performance with respect to both ESPRIT and Smyth algorithms at low SNR can be obtained by a regularized filtering procedure on the data.