Hankel Range and Null Space Methods for Harmonic Retrieval in Unknown Colored Noise

The utility of the Hankel covariance matrix in estimating the frequencies of sinusoids corrupted by unknown colored noise is evaluated. Specifically, we address, in a comparative setting, some range and null space methods, including the use of nominally excessive matrix dimensions, higher order covariance lags, and linear combinations of null space elgenvectors in a root locus setting. Numerical results suggest that while range and null space methods can yield comparable results, the former are more sensitive to retaining low lags. Methods which take greater advantage of theoretical redundancy are also addressed.