A higher-order statistics-based phase identification of three-component seismograms in a redundant wavelet transform domain

Abstract

An efficient identification of P and S phases of three component seismograms (3-CS) using higher-order statistics (HOS) in a redundant wavelet transform domain (R-WT) is presented in this work. The R-WT peals the 3-CS signal into several frequency layers, where the HOS-based criteria (cross-cumulants, skewness, kurtosis) could reflect efficiently the onset times of the P and S phases of an event, during 3-CS recordings. This is of great importance to seismologists, since, afterwards, the epicenter of an earthquake could be accurately identified. The implementation of this approach on pre-classified 3-CS, recorded during several events from seismological stations was examined. Experiments have shown that a promising and automated estimation of the P and S phase onset times, consistent with that of the analysis, can be achieved even in noisy cases (e.g. additive Gaussian noise), due to properties of HOS.