APPLICATIONS OF PARAMETRIC AND NON-PARAMETRIC SPECTRAL ESTIMATION TECHNIQUES TO PASSIVE SONAR DATA.

Methods of detection, classification, localisation and tracking using sonar data rely heavily on the use of Fourier techniques. Examples of non-parametric methods for estimating the power spectrum of single hydrophone data in both stationary and non-stationary environments are given. Some improvements resulting from the use of parametric methods (e. g. , maximum entropy) are presented. The role of the cross-power spectral matrix of receiver outputs from hydrophone arrays in estimating the power spectrum as a function of angle or wavenumber is discussed. Use of the cross-power spectral matrix and a spatial covariance function derived from it allows the application of parametric methods (e. g. , maximum likelihood or maximum entropy) to the analysis of data from arrays of receivers.