Evaluation of an autoregressive spectral estimation technique for determining horizontal wave-number content in shallow water

Abstract

Modal-based techniques for geoacoustic inversion require estimates of discrete horizontal wave numbers corresponding to the propagating modes in a shallow-water waveguide. For range-dependent waveguides, local wave-number content can be extracted as a function of range using methods analogous to short-time Fourier transform techniques as applied to non-stationary time series data. A limiting factor in these techniques is the aperture length, L, required to resolve closely spaced horizontal wave-number content. Using classical spectral estimation methods, the limit of resolution is proportional to 1/L. In order to increase the spatial resolution of local horizontal wave-number estimates, the use of a high-resolution autoregressive spectral estimator was investigated for determining horizontal wave-number content. The performance of the estimator was evaluated numerically using both time-series and spatial data. Results are presented of estimator performance for various aperture lengths using signals with increasing levels of additive noise and signals with closely spaced wave-number components. Comparisons are made with results of classical methods as well as other high-resolution methods. The results show the AR method to be well suited for wave-number estimation in shallow-water waveguides. [Work supported by ONR.] a)Presently with the Acoustics Group, The Pennsylvania State Univ., Appl. Res. Lab., P.O. Box 30, State College, PA 16804, kmb166@psu.edu.