Earthquake epicenter location using time-frequency representations

Epicenter location can be determined using the times of arrival of the P and S phases present in the seismogram. Different methods have been applied for seismic detection. The best results have been obtained when multi-station three-dimensional data are used, but a reliable phase detection technique using one-dimensional data does not exist. Time-Frequency Representations (TFRs) are an alternative tool to process the seismic trace and to detect the P and S waves because the seismogram is a non-stationary signal. However, the usefulness of TFR analysis depends largely on the transformation kernel. In this work we propose a TFR-based algorithm using two types of kernel: a Line Dirac-delta kernel and the magnitude of the ambiguity function of a low noise P wave. Results of experiments with real seismograms show that not only is it possible to detect the P and the S phases, but also the technique provides frequency characterization that can be used to distinguish between both waves even in cases where the seismogram is very noisy