High resolution feature extraction from reverberation data

Abstract

A new method of signal approximatzon is presented which has the capability to model complex active returns due to reverberation. The method explored herein is data adaptive in the sense that high resolution processing is attempted where necessary, while fast algorithms are used to estimate the remaining components. The basic goal is to achieve accurate and unbiased estimation of the delays and time scalings of a group of closely spaced signal returns. A secondary goal is to achieve this estimation performance without sacrificing the cornputatzonal eficiency of the fast algorithms available.