Short-time least squares spectral analysis of pass-by noise in water from a rigid inflatable boat

Abstract

This paper investigates least squares spectral analysis as a tool to analyze non-stationary signals from pass-by noise measurements in water. The spectral analysis involves successive least squares fitting of a finite Fourier series to approximate the observation in a piecewise manner. The least squares spectral analysis is used to search the signals for first- and second-order periodicity as well as the presence of fundamental periodicity. A first-order analysis reveals line components in the signals, whereas a second-order analysis reveals periodic amplitude modulations. Analysis with a higher-order finite Fourier series reveals harmonic structures in the signals. The main contribution of this paper is the model of a magnitude-squared cosine wave which can be used to analyze second-order periodicity. The developed short-time least squares spectral analysis is illustrated on noise radiated from a rigid inflatable boat in shallow water.