Analysis of a direct separation method based on adaptive chirplet transform for signals with crossover instantaneous frequencies (Q2105101)

How to retrieve the modes (also called components or sub-signal building blocks) of a multi-component signal accurately is an important issue in applications. Empirical mode decomposition (EMD), synchrosqueezing transform (SST), signal separation operation (SSO), and iterative filtering decomposition (IFD) have been proposed and developed for signals with a well-separation of the sub-signal frequency curves. Recently, the authors [J. Comput. Appl. Math. 396, Article ID 113607, 19 p. (2021; Zbl 1465.42008)] have proposed the chirplet transform-based signal separation scheme (CT3S) to recover modes for signals with crossover instantaneous frequencies. CT3S extends SSO from the two-dimensional ``time-frequency'' plane to the three-dimensional ``time frequency-chirp rate'' space to remove the frequency-separation specification. In this paper, the authors focus on carrying out an in-depth error analysis of instantaneous frequency (IF) estimation and component recovery of the CT3S method. Here, the error bounds for IF estimation and component recovery are established for the admissible window function, and explicit expressions for these bounds are provided when the window function is the Gaussian function. It should be pointed out here that the authors propose the time-scale-chirp rate transform, based on adaptive continuous wavelet transform (CWT) for recovery of non-stationary signal components with crossover instantaneous frequency curves before. These direct separation methods are significant in theory and applications for multi-component signals.