Decomposition of non-stationary signals into varying time scales: Some aspects of the EMD and HVD methods

The Empirical Mode Decomposition has become very popular since its first introduction. Its suitability and expected performance for specific signal processing task is however somewhat open ended.Addressed are basic questions concerning the decomposition of signals according to different time scales, from noise sensitivity to frequency resolution. Comparing it to the recently introduced Hilbert Vibration Decomposition, signals simulating moving mechanical systems are analyzed. It is argued that in spite of signal adaptive properties, a universal fully automatic decomposition is still beyond our capabilities.

► Concerning decomposition of non-stationary signals with different time scales. ► Presenting examples of noise sensitivity, frequency resolution, and noise. ► Comparing SFFT, EMD, and HVD, signals simulating mechanical systems are analyzed. ► For adaptive procedures an automatic decomposition is still beyond our capabilities. ► Results need to be checked and always compared with a physical interpretation.