ИСТИНА |
Войти в систему Регистрация |
|
ИПМех РАН |
||
This poster is devoted to reliable automatic detection of seismic phases, an important problem in seismometry related to many aspects of seismological theory and practice. Most seismograph networks dispersed over the globe lose much in terms of event location accuracy and event classification, when the noise exceeds a certain threshold, even though at a single station only. The system we are developing can help toward solving that problem. Its main functions include the classification of seismic events and the identification of low energy phases at high levels of seismic noise. The noise in question may be anything from natural to man-induced and from stationary to impulsive. The system is being developed for early warning about an earthquake that has occurred and is based on the single seismometer principle. Now because the system is supposed to be installed at the end user's, that is to say, in locations subject to increased levels of manmade noise, we focus our study on signal detection precisely under these conditions. The system is implemented to include three detection processes in parallel. The first is based on the study of the co-occurrence matrix of the signal wavelet transform. The second consists in using the method of a change point in a random process and signal detection in a moving time window. The third uses artificial neural networks as a versatile classifier. Further, applying a decision rule we carry out the final detection and estimate its reliability. We discuss the history of the issue and a possible block diagram for the device in question, present some results obtained on test examples, as well as consider whether the system can be used for reliable detection and classification of seismic events by stations in seismograph networks.