Comparison of time-frequency representations of random wave elevation data

This paper is concerned with examining the nonstationary characteristics of laboratory-generated random waves. To do this we utilize and compare a number of time-frequency distributions belonging to Cohen's class, namely, the spectrogram, and the Wigner-Ville, Choi-Williams, and the reduced-interference distribution (RID). We also consider time-scale auto-power spectra, cross-spectra, and linear coherence spectra all of which are based on the continuous wavelet transform. With regard to Cohen's class, the RID distribution appears to be most suitable for analyzing random wave data. The wavelet-based approach appears to be most appropriate for generating complex valued time-scale spectra, such as the cross-power spectrum and the bispectrum.