DISCRIMINATION OF NATURAL EARTHQUAKES FROM ARTIFICIAL EXPLOSIONS USING A NEW FLVQ MODEL

In this paper, a new clustering technique is used for seismic discrimination purposes, based on the P-wave spectra computed from the short period teleseismic recordings. In this study, we have examined the different schemes of LVQ (Learning Vector Quantization) methods for clustering the six spectral features, extracted from the seismic signals. The conventional LVQ with Kohonen learning rule is first studied. A new type of FLVQ has been proposed on extension and generalization of the pervious scheme presented by Sakuraba (۱۹۹۱). Another new algorithm has also been proposed to calculate the optimum number of clusters in each of the feature planes. The two new algorithms are used as nested sub-algorithms to make up the learning phase. A comparison has been made among the above mentioned clustering methods, using the leave-one-out testing strategy