Adaptive synchrosqueezing transform with a time-varying parameter for non-stationary signal separation (Q2197960)

A variety of signals of practical interest, are multicomponent signals, where each component follows some characteristic amplitude and frequency (or phase) modulation laws. The analysis of such signals involves estimation of the instantaneous amplitude and frequency of each component of the signal. The instantaneous frequency (IF) of a signal/component of a signal is a parameter which is often of significant practical importance. In this study authors indicate the possibility of a parametric analysis and estimation of IF of multicomponent transient signals of real world, using the adaptive CWT, the adaptive SST and the 2nd-order adaptive SST, all with time-varying parameter for the commonly used wavelets A detailed mathematical derivation and performance assessment of their algorithm is presented illustrating with the plots of the results obtained from scientific computing, Bump wavelet and Morlet wavelet are used as reference (benchmark) wavelets for the theoretical development. As described, synchro squeezing uses the continuous wavelet transform (CWT) and its first derivative with respect to translation. The CWT is invertible and since the SST inherits the CWT invertibility property, the signal can be reconstructed. SST has been proved to be well adapted to multicomponent signals in terms of separability of components and at the same time achieving optimal accuracy, overcoming in some way the limitations imposed by the uncertainty principle on the simultaneous resolution in time (position) and frequency. The new method proposed by the authors turns out to be more efficient in separating components close to each other than the conventional EMD and SST as demonstrated in the empirical results. To drive home the point, about the efficacy of the method proposed, research findings are reported with computer generated plots as an aid to visualization. The underlying strategy of the proposed method is as follows: \par 1) Obtaining the condition under which a separable time-varying parameter can be selected such that the corresponding adaptive SST of individual components do not overlap. \par 2) Deriving formula and proposing an algorithm to select optimal value of the parameter such that the adaptive SST of each component will have sharper representation and better IF estimation \par 3) In addition, the fulfillment of the above condition ensures that the adaptive SST satisfies the energy conservation property. In the concluding section, experiments on signals with more than two components are conducted to establish the efficiency, robustness and versatility of the proposed method. Summing up, in this paper, the estimation of IF of nonstationary multicomponent signals is addressed. The CWT-based SST (WSST) works through squeezing the CWT, where the wavelet is required to be admissible. The method and algorithm proposed in this paper can be applied to the case of the STFT-based SST as well. Another significant contributions to TF analysis which merits mention is the second order adaptive SST introduced by the authors differs from the one described in the earlier published work in approach and performance and is capable of generalization. Comparison experimental results help to establish the superiority and competitiveness of the proposed adaptive CWT and adaptive SST in separating components of multicomponent signals with fast varying frequencies. To make the article as self-contained as possible, proofs of propositions and theorems are provided in Appendix.