New heuristics based on wavelet analysis of a single sensor record for earthquake and explosion detection
Federal Research Center “Unified Geophysical Survey of the Russian Academy of Sciences”
Journal: Seismic instruments
Tome: 58
Number: 3
Year: 2022
Pages: 5-24
UDK: 550.(34.(012:062):344.2)
DOI: 10.21455/si2022.3-1
Show citation
Silkin K.Yu. New heuristics based on wavelet analysis of a single sensor record for earthquake and explosion detection
// . 2022. Т. 58. № 3. С. 5-24. DOI: 10.21455/si2022.3-1
@article{SilkinNew2022,
author = "Silkin, K. Yu.",
title = "New heuristics based on wavelet analysis of a single sensor record for earthquake and explosion detection
",
journal = "Seismic instruments",
year = 2022,
volume = "58",
number = "3",
pages = "5-24",
doi = "10.21455/si2022.3-1",
language = "English"
}
Copy link
Copy BibTex
Keywords: recognition, earthquake, explosion, underground explosion, career explosion, wavelet transformation, correlation
Аnnotation: Recognition of the seismic event by the type of its phenomenon (earthquake or explosion, and if the explosion, then an underground explosion or an explosion in a career) at a regional scale on its seismogram is the task that many researchers are solved throughout the world. The widescreen detailed review of Russian and global publications on this topic have been produced. This review made it possible to formulate the most promising directions on which research is underway. Thus, this work offering another approach to creating a discriminatory feature may be useful to improve the result of the recognition of the seismic event. The basis of the proposed method is the continuous wavelet analysis of the seismogram of a single receiver. Two additional transformations (conducting frequency envelopes to waveletogram and their crosscorrelation at a given time) sequentially translate this result in compact frequency-time portrait of the event. Approbation of this technique was carried out on seismograms of several events, the nature of which is a priori known. Recognition is possible both visually (including machine vision methods) and automatically. For the first option, the key features of frequency-temporal portraits of events that should be paid attention to are formulated. For the second case, a method for determining the numerical characteristics measured by the obtained images is defined. It is shown that these characteristics are naturally divided into clusters that correspond to the nature of the events.