Evaluation of high-resolution frequency estimation methods for determining frequencies of eigenmodes in shallow water acoustic field

Abstract

In shallow water, the acoustic field in the far field, due to a point source, can be modeled as a sum of contributions from trapped modes propagating in the wave guide. In many applications, it is necessary to estimate the eigenvalues of these modes from a measurement of the acoustic field made on a horizontal array using a monochromatic source. In this paper the performance of two high-resolution methods (MUSIC and ESPRIT) in estimating the eigenvalues of the modes from the measured field is evaluated. In particular, we investigate the ability of these methods to estimate the frequencies accurately for various signal to noise ratios and their ability to resolve closely spaced frequencies using synthetic noisy data. Of the two methods, ESPRIT had the better performance in terms of the variance of the estimates. Simulation performed to study the effect of modeling errors on the performance of the algorithms showed that, for a medium that is weakly range dependent, the performance of the algorithms is not adversely affected. But errors in ranging and the presence of colored noise in data have considerable impact on their performance. The MUSIC algorithm shows marked degradation in performance when there are ranging errors. Taken overall, ESPRIT has the most acceptable performance levels, both in terms of its accuracy of the estimates and its robustness in the presence of modeling errors.