Comparison of time-frequency representations of signals exhibiting spectral discontinuities

Short-time-Fourier-transform (STFT), pseudo-Wigner-Ville, constant-coefficient autoregressive parametric and time-dependent-coefficient ('evolutionary') autoregressive parametric representations of nonstationary random signals with jumps are compared in simulations using synthetic signals. The methods and simulation techniques are described in detail, and the results are presented in graphs and photographs of the synthetic spectra and their representations. STFT and constant-coefficient parametric methods are found to give better results under the conditions studied.