Information preserving empirical mode decomposition for filtering field potentials

This paper presents a concept of using the empirical mode decomposition (EMD) as a filtering tool to extract information-preserving intrinsic mode functions (IMFs) in an adaptive manner. The approach is tested on several local field potentials (LFPs) and information quantification is carried out in spectral domain using Shannon's Information. The study suggests that not all IMFs are information carriers. It is found that the 1st IMF carries 60-80% of the total information from original LFP and few informative IMFs are usually the main information carriers. Adding more IMFs does not increase the information level. For different datasets, the order of the informative IMFs varies and by using information preserving EMD, only few IMFs are retained to provide a simplified representation of underlying oscillations contained in LFPs. © 2009 Springer Berlin Heidelberg.