Acoustic and seismic time series analysis using ensemble empirical mode decomposition

Abstract

Empirical mode decomposition (EMD) is an adaptive or data-driven time series analysis technique ideally suited to investigate non-stationary signals. EMD decomposes the signal into fast and slow oscillations called intrinsic mode functions (IMFs). Ensemble EMD (EEMD) methods have been developed to alleviate the mode-mixing phenomenon present in the EMD technique. Various real signals will be analyzed using these methods to investigate whether the IMFs correlate to acoustic modes. These real signals include underwater acoustic signals from broadband sources, Scholte and Rayleigh wave signals and music signals. These underwater acoustic signals and the seismic signals exhibit multi-modal structure. In addition these modes are also dispersive in nature. Accurate resolution of these modes in the time-frequency plane is critical for the estimation of medium properties via inverse schemes. Finally, the EEMD technique will be used to investigate the complex dynamic structure of the pitching structure in South Indian Classical music. In all these cases investigated the mapping of the signal into the time-frequency plane reveal distinct features representing the modal dispersion and pitching structure, respectively. EEMD will be used to extract these features accurately with high resolution. [Work supported by Office of Naval Research.]