Analysis and classification of time-varying signals with multipletime-frequency structures

摘要

We propose a time-frequency (TF) technique designed to matchnsignals with multiple and different characteristics for successfulnanalysis and classification. The method uses a modified matching pursuitnsignal decomposition incorporating signal-matched dictionaries. Fornanalysis, it uses a combination of TF representations chosen adaptivelynto provide a concentrated representation for each selected signalncomponent. Thus, it exhibits maximum concentration while reducing crossnterms for the difficult analysis case of multicomponent signals ofndissimilar linear and nonlinear TF structures. For classification, thisntechnique may provide the instantaneous frequency of signal componentsnas well as estimates of their relevant parameters