Applications of Bispectral Analysis.

Estimates of the power spectral density have been found to be very useful in a variety of signal processing applications over the last several decades. Higher order spectral contain information not present in the power spectrum and recently, estimates of higher order spectral have been shown to be useful in certain signal processing problems. In particular, estimates of the bispectrum and bicoherence have been found useful in detecting non-Gaussianity and non-linearity, in system identification, and in detecting transient signals. This paper is concerned with the estimation of the bispectrum and bicoherence of underwater acoustic signals. We first introduce and describe some of the properties of the bispectrum and bicoherence. The bispectral estimation algorithm as implemented here at the Marine Physical Laboratory is then described in detail. Finally, several bispectra are estimated from actual data and the results are analyzed. In the results portion of the paper we concentrate on three particular properties of bispectrum estimation. We show how the bispectrum estimate can be used to detect non-Gaussianity, non-linearity, and harmonic coupling. The detection of harmonic coupling is shown to be a useful property of the particular bispectrum estimation algorithm employed here. AN