Dynamic Bayesian Networks for Real-Time Classification of Seismic Signals

Abstract

Abstract. We present a novel method for automatic classification of seismological data streams, focusing on the detection of earthquake signals. We consider the approach as being a first step towards a generic method that provides for classifying a broad range of seismic patterns by modeling the interrelationships between essential features of seismograms in a graphical model. Through a continuous Wavelet transform the features are extracted, yielding a time-frequencyamplitude decomposition. The extracted features obey certain Markov properties, which allows us to form a joint distribution in terms of a Dynamic Bayesian Network. We performed experiments using real seismic data recorded at different stations in the European Broadband Network, for which we achieve an average classification accuracy of 95%.

Characterising Seismic Data

“…Application-wise, the Dynamic Bayesian Network (DBN) approach to seismic data classification in [6] is closely related to our research. They also decompose the signal using wavelets -albeit a different one: the Morlet wavelet -and then use a DBN to classify the incoming data as either "Earthquake" or "Noise".…”

“…Fortunately, it is well known in seismology that the signal in the range from roughly 4 Hz to 1 /4 Hz has a fairly good signal to noise ratio [1,6]. That is, signals in that range are predominately of seismic origin.…”

“…For example, assume that we have a symbol a on level l at time t. To predict a, Physics tells us [6] we can use a "block" of symbols on all levels at all time points prior to t provided this block contains no holes and it is adjacent to a. Clearly these patterns satisfy an Apriori property and thus can be found with a standard levelwise algorithm.…”

The Automated Identification of Volcanic Earthquakes: Concepts, Applications and Challenges

“…Other ones, such as the Hilbert and wavelet transforms have been applied for representation; e.g. by Riggelsen et al (2007), San-Martín et al (2010), andPorro-Muñoz et al (2010b).…”

“…A generalization of HMMs are the so-called dynamic Bayesian networks. Riggelsen et al (2007) Duin et al (2010). Authors of the first study built classifiers on top of a classical feature representation while the others employed simple ones, either in the dissimilarity space or to be combined in a second step of the classification process as explained at the end of Sec.…”