Search

The seismic P phase first arrival identification is a fundamental problem in seismology. The accurate identification of the P-wave first arrival is not a trivial process, particularly when the seismograms present a very low signal-to-noise ratio (SNR) or are contaminated with artificial transients that could produce false alarms. In this paper, a new approach based on higher order statistics and the stationary wavelet transform is presented. The P onset is obtained under a statistical criterion applied in the time--frequency domain. The results have been compared to those estimated by another P phase picking algorithm and P onsets picked by expert analysts. The comparison shows that our proposed method efficiently provides a good estimate of the P onset picks that are consistent with analyst picks, particularly in cases of very low SNR. Index Terms--Kurtosis, P phase identification, seismic signal processing, stationary wavelet transform (SWT).