Seismic event discrimination using neural networks

To date , the problem or seismic event detection and source location has been well documented, bu t the task of discriminating between diflerent seismic sources has typically used heuristic, ofl-line methods. I n this paper we present a method of automatically discriminating between seismic events by using a multilayer perceptron neural network system. T h e seismic signals of concern in this research are those generated by earthquakes and underground nuclear explosions. T h e motivation for this research is the idea of using seismic information to verify compliance with respect to a Comprehensive Tes t Ban Treaty. A simple 2-l%yer network correctly classified 11 ou t of 12 events in a d a t a tes t set.