A decision support system for classification and recognition of earthquakes and explosions

This paper introduces current advances and some rudimental results of our ongoing research project. To discriminate between earthquakes and explosions, temporal and spectral features extracted from seismic waves, plus some seismological parameters (such as epicenter depth, location, magnitude) are crux for rapid and correct recognizing event sources (earthquakes or explosions). Seismological parameters are used as the first step to screen out obvious earthquake events. Fourier transforms (FFT), chirp-Z transforms, wavelet transforms have been conducted and some prominent features are acquired by present experimental dataset. In some experiments, wavelet features plus support vector classification (SVC) have reached very high correct recognition rate (>95%). If more temporal and spectral features are utilized properly, more robust recognition result is surely possibility.