Application of time-frequency analysis to non-stationary and multicomponent signals

In this paper, four different approaches to time-frequency analysis performed by using parametrical and non-parametrical bispectral estimation, as well as the Wigner-Ville and Wigner-bispectrum techniques are considered, studied and compared between each other. Frequency resolution and noise immunity have been investigated for different time-frequency distributions by the computer simulations. Results of computer simulations are represented both for multi-component signal models and radar backscattering signals experimentally recorded for moving radar object in the form of walking human. It is demonstrated that the parametrical bispectrum-based technique provides smallest distortions in time-frequency distributions for the non-stationary multicomponent signals though, at the same time, it inferiors a few to the Wigner-Ville distribution in noise immunity. © 2017 by Begell House, Inc.