Caracterización de señales sísmicas utilizando modelos paramétricos y transformada cepstrum

Abstract

This work proposes a methodology for the extraction of features from seismic signals that allows to identify different types of volcanic earthquakes that have been studied in the Volcanological and Seismological Observatory of Manizales (OVSM).This process is made through a joint representation on time and frequency known as adjustment surface. Autoregressive parametric models (AR, ARMA) and the cepstrum transform are compared on the identification of random processes. Different criteria for the selection of the parametric models order are used and the effect of some normalization methods over the characteristicsis studied. Recognition rates of 99.8% between two kinds of volcanic signals are obtained.