AN EXAMINATION OF SOME NEW AND CLASSICAL SHORT PERIOD DISCRIMINANTS

摘要

Optimum linear and quadratic discrimination filtering techniques are developed for discriminating between short period seismic records originating from earthquakes and explosions. Linear and quadratic detection filtering and matched filtering are compared with the classical spectral ratio and complexity measure using a learing population of LASA array beams of 23 earthquakes and 15 explosions and a test population with 17 earthquakes and 11 explosions. The linear detection filter misclassifies one event in the test set whereas all other techniques misclassify between three and five events. It is shown that for the spectral ratio discriminant the discriminatory power lies in the ratio of .4-.8 Hz energy to 1.0 Hz energy, and that the higher frequency energy has no additional discriminatory power.