Maximum Error in Discrete EMD Decomposition of Periodic Signals

The present investigation concerns the recently developed Hilbert-Huang transformation. This technique is expected to decompose time-dependant data series into its individual characteristic oscillations with the so-called empirical mode decomposition (EMD). EMD is capable to decompose empirically any complex set of data into a finite number of Intrinsic Mode Functions (IMF). The aim of the present experimental study is to contribute to a better understanding of particular aspect of EMD. We consider the spectral analysis of the IMF components of digital signals. We show that errors of EMD decomposition of sine wave are in fact new frequency components called artefacts. These artefacts depend on a signal frequency and sampling frequency. In this paper, artefacts are explored and maximum error is analysed in spectral domain.