A New Intrinsic Mode Function Selection Method Based on Power Spectral Density

Abstract

Emprical Mode Decomposition is one of the mostly used decomposition method to decompose non-linear and non stationary multivariate signals into linear and stationary sub signals. Intrinsic mode functions, which are the sub-signals obtained as an output of the process are used in many studies, primarily classification. Even though, some of the intrinsic mode functions are noise dominant while some of them are signal dominant. For this reason and that to reduce the computational cost, intrinsic mode function selection methods have been proposed in the past. In this study, a new selection method is proposed which is inspired of proposed methods in the past.