Complex Singular Spectrum Analysis Leveraging Adaptive Taper Windows for Enhancing Mode Reconstruction From Multivariate Signals

Abstract

In this letter, a generic extension of complex singular spectrum analysis (CSSA), referred to as GC-SSA, is proposed to enhance mode reconstruction from multivariate signals. This is achieved by introducing adaptive taper windows for CSSA. Specifically, we formulate an optimization problem related to window design for specific multivariate signals, and then employ an iterative algorithm to optimize the coefficients of the taper windows. GC-SSA using optimized taper windows can decompose multivariate signals and perfectly reconstruct time-varying modes that have maximally concentrated energy. Numerical simulations were demonstrated to validate the effectiveness of the proposed method in mode reconstruction compared to other multivariate signal processing methods.