Neural network approach to seismic signal analysis

Abstract

This paper summarizes the preliminary findings of a Neural Network approach to automatic classification of moving objects based on seismic signals detected through a geophone. Four types of moving objects were considered: human beings, motorycles, cars and buses. A 32-16-4 network structure was used and the data was preprocessed before neural analysis. The results reveal 92% accuracy in classifying human beings from vehicles. However, only 74% accuracy was achieved in classifying the four different types of moving objects.