AN EFFICIENT ALGORITHM FOR 2-DIMENSIONAL FREQUENCY ESTIMATION

This paper presents a computationally efficient eigenstructure-based 2 D-MODE algorithm for two-dimensional frequency estimation. We derive t he theoretical performance of the 2D-MODE estimator and show that it i s asymptotically statistically efficient under either the assumption t hat the number of temporal snapshots is large or the signal-to-noise r atio is high. Numerical examples showing the performance of this algor ithm and comparing it with the computationally efficient subspace rota tion algorithms are also given. We show that the statistical performan ce of the 2D-MODE algorithm is better than that of the subspace rotati on methods. The amount of computations required by the former is no mo re than a few times of that needed by the latter for either small numb ers of spatial measurements or a single temporal snapshot, which are t he cases of interest herein.