Empirical Mode Decomposition: Theory and Applications in Underwater Acoustics

Abstract

Empirical mode decomposition (EMD) is a signal processing method that produces a data-driven time-frequency representation suited to characterize time-varying and nonlinear phenomena. In EMD, intrinsic mode functions (IMF) are sequentially estimated from the signal of interest to represent different intrinsic oscilation modes and produce an orthogonal representation of the original information. Different algorithms have been proposed for EMD estimation to deal with limitations such as mode-mixing and noise sensitivity. To obtain a frequency-domain representation, EMD is usually associated with the Hilbert transform, in this case, the method is referred to as the Hilbert-Huang transform (HHT). This paper presents a theoretical review of the fundamental aspects of both EMD and HHT, such as IMF estimation procedure and IMF orthogonality. Variations of the original EMD algorithm are also presented. Both simulated and experimental underwater acoustic signals are used to illustrate the efficiency of EMD/HHT in revealing relevant characteristics from time-varying and nonlinear information.