A Sparse TFD Reconstruction Approach Using the S-method and Local Entropies Information

Abstract

This paper aims to investigate the S-method (SM) as an alternative for the Wigner-Ville Distribution (WVD) when used as the starting point for a sparse time-frequency distribution (TFD) reconstruction of non-stationary signals. The motivation comes from the SM's ability of providing a high-resolution TFD with satisfactory cross- and inner-artefact suppression, which should lead to a reconstructed TFD performance improvement over the WVD. The comparison between the WVD and the SM has been conducted using several state-of-the-art algorithms optimized with the multi-objective meta-heuristic optimization method (by minimizing the mean squared error between the local number of components in the starting and reconstructed TFDs and the number of regions with continuously connected samples). The results are shown for single and multi-component noisy synthetic signals.