Empirical mode decomposition in a time-scale framework.

The analysis of multicomponent signals, made of a small number of amplitude modulated - frequency modulated components that are overlapped in time and frequency, has gained considerable attention in the past years. These signals are often analyzed via Continuous Wavelet Transform (CWT) looking for ridges the components generate on it. In this approach one ridge is equivalent for one mode. The Empirical Mode Decomposition (EMD) is a data-driven method which can separate a signal into components ideally made of several ridges. Unfortunately EMD is defined as an algorithm output, with no analytical definition. It is our purpose to merge the data-driven nature of EMD with the CWT, performing an adaptive signal decomposition in a time-scale framework. We give here a new mode definition, and develop a new mode extraction algorithm. Two artificial signals are analyzed, and results are compared with those of synchrosqueezing ridge-based decomposition, showing advantages for our proposal.