Spectrum Estimation Techniques for Characterization and Development of WT4 Waveguide - 1

The problem of estimating the spectrum of a stationary time series has appeared frequently in the scientific literature and myriad approaches have been suggested. Nonetheless it became apparent during the course of the development of the WT4 waveguide system that these methods were inadequate for many of the data sets of interest. The techniques presented here were therefore developed. It is commonly stated that the method selected to estimate a spectrum depends on the ultimate use of the estimate, and unfortunately to some extent this is true. The method described below is felt to represent an advance in that the basic technique works well in a variety of cases which previously would have required individual treatment. The loss calculations reported in Anderson et al.1 are indicative of its accuracy. The procedure which has evolved for estimating spectra can best be described as robust adaptive prewhitening. Such methods have three distinct stages: formation of a pilot spectrum estimate, using this estimate to design a prewhitening filter, and finally giving the result as the ratio of the spectrum of the filtered data to the power transfer function of the filter. This method is potentially both efficient and robust. The 1769