Seismic signals detection and classification using artiricial neural networks

Abstract

Pattern recognition belongs to a class of Problems which are easily solved by humans, but difficult for computers. It is sometimes difficult to formalize a problem which a human operator can casily understand by using examples. Neural networks are useful in solving this kind of problem. A neural network may, under certain conditions, simulate a well trained human operator in recognizing different types of earthquakes or in detecting the presence of a seismic event. It is then shown how a fully connected multi layer perceptron may perform a recognition task. It is shown how a self training auto associative neural network may detect an earthquake occurrence analysing the change in signal characteristics.