Iterated SVD for Improving Spectral Resolution of Nonstationary Signal

Abstract

It is well known and proved already that principal singular vectors of Hankel data matrix repeats same spectral structure and by concatenation, these singular vectors allows us to gain high statistical stability and at the same time improve spectral resolution, but the main question arise; are they working well when frequencies are separated well enough or can they detect hidden periodicities even behind the resolution limit? In this article iterated SVD method is used to increase length of the time series much bigger than single SVD can do and at the same time increase spectral resolution and decrease noise in a new time series at more level then only one SVD can do.