Seismic Signal Classification using Multi-Layer Perceptron Neural Network

The aim of the present study is to investigate and explore the capability of the multilayer perceptron neural network to classify seismic signals recorded by the local seismic network of Agadir (M orocco). The problem is divided into two main steps, the feature extraction step and classification step . In the former, relevant discriminant features are extracted from the seismic signal based on the time and frequency domains. These are selected based on the analysts' experience. In the latter step, a process of trial an error was carried out to find the best neural network architecture. Classification results on a data set of 343 seis mic signals have demonstrated that the accuracy of the proposed classier can achieve more than 94%.