Abstract—The automatic discrimination of seismic signals is an important practical goal for earth-science observatories due to the large amount of information that they receive continuously. An essential discrimination task is to allocate the incoming signal to a group associated with the kind of physical phenomena producing it. In this paper, two classes of seismic signals recorded routinely in geophysical laboratory of the National Center for Scientific and Technical Research in Morocco are considered. They correspond to signals associated to local earthquakes and chemical explosions. The approach adopted for the development of an automatic discrimination system is a modular system composed by three blocs: 1) Representation, 2) Dimensionality reduction and 3) Classification. The originality of our work consists in the use of a new wavelet called "modified Mexican hat wavelet" in the representation stage. For the dimensionality reduction, we propose a new algorithm based on the random projection and the principal component analysis. Keywords—Seismic signals, Wavelets, Dimensionality reduction, Artificial neural networks, Classification. I. INTRODUCTION S the earthquakes, a chemical explosion, an underground nuclear explosion, a volcanic eruption or generally any event that can generate vibrations of the soil creates seismic signals that propagate in the ground. In order to be a monitoring tool, a seismic network must be able to identify the source of different seismic signals. This task consists to discriminate between different events: local earthquake, regional earthquake, chemical explosion, nuclear explosion, etc. The manual discrimination of digital records is a difficult task that demands considerable efforts and costs. In this sense, several works have been developed to automate the discrimination task, we note: [1], [2], [3]. In this paper, we present a modular system for the classification of seismic signals. Three blocs compose this system (see Fig. 1): Representation, Dimensionality This work was supported by PROTARS III (CNRST, Rabat, Morocco) under Grant D48. F. M. BENBRAHIM is with the Ecole Mohammadia d'Ingénieurs, Rabat, Morocco (corresponding author to provide phone: 212-63631938; fax: 212- 55633878; e-mail: benbrahim10@yahoo.fr). S. A. DAOUDI is with the Régie Autonome de Distribution de l’eau et de l’électricity de Marrakech, Marrakech, Morocco (e-mail: adil75@menara.ma). T. K. BENJELLOUN is with the Ecole Mohammadia d'Ingénieurs, Rabat, Morocco (e-mail: bkhalid@emi.ac.ma). F. A. IBENBRAHIM is with the Centre National pour la Recherche Scientifique et Technique Rabat, Morocco (e-mail: ibenbrahimr@cnr.ac.ma). reduction, Classification. The advantage of this system is the ability to profit from the existent evolutions of each module in any time. FIGURE 1: SYSTEM DIAGRAM The remainder of this paper is structured as follows: first, we present the different representations for seismic signals and we propose a new complex wavelet. We then describe our algorithm for the part dimensionality reduction. The next paragraph is dedicated to classification by multilayer perceptron network. Finally, we discuss the results and the conclusions. II. SEISMIC SIGNALS REPRESENTATIONS The choice of the representation is an important parameter that it must be chosen carefully in order to increase the classification performances. It consists to define a representation space permitting the extraction of the pertinent information. For seismic signals, all previous works have highlighted that a representation space, where the power, the frequency and the time are .present, is an adequate space. We can replace the frequency parameter by the scale parameter that permits a multiresolution analysis of the signal. A. Time and Frequency Representations The time representation is the natural form to represent a signal deriving from a given phenomena. It is not necessary to use any mathematical tool to perform it and we can have some information about signal from it. However, it is not adapted for the automatic classification. The frequency representation is an alternative for the temporal representation of looking at a signal. It consists to represent the frequency content of the signal via the Fourier transform. In the seismic signals case, the frequency contents and all the statistic properties change with the time. Consequently, for the events with weak signal noise ratio, the classification based on the Fourier transform can give wrong results. Moreover, this representation limits the generalization Discrimination of Seismic Signals Using Artificial Neural Networks Mohammed BENBRAHIM, Adil DAOUDI, Khalid BENJELLOUN, and Aomar IBENBRAHIM A Seismic Data Representation Dimensionality Reduction Classification World Academy of Science, Engineering and Technology International Journal of Computer and Information Engineering Vol:1, No:4, 2007 984 International Scholarly and Scientific Research & Innovation 1(4) 2007 ISNI:0000000091950263 Open Science Index, Computer and Information Engineering Vol:1, No:4, 2007 publications.waset.org/10314/pdf