Technique of Informative Features Selection in Geoacoustic Emission Signals

Abstract

Studies of geoacoustic emission in a seismically active region in Kamchatka show that geoacoustic signals produce pronounced pulse anomalies during the earthquake preparation and post-seismic relaxation of the local stresses field at the observation point. The qualitative selection of such anomalies is complicated by a strong distortion and weakening of the signal amplitude. A review of existing acoustic emission analysis methods shows that most often researchers turn to the analysis of more accessible to study statistical properties and energy of signals. The distinctive features of the approach proposed by the authors are the extraction of informative features based on the analysis of time and frequency-time structures of geoacoustic signals and the description of various forms of recognizable pulses by a limited pattern set. This study opens up new ideas to develop methods for detecting anomalous behavior of geoacoustic signals, including anomalies before earthquakes.
 The paper describes a technique of information extraction from geoacoustic emission pulse streams of sound frequency range. A geoacoustic pulse mathematical model, reflecting the signal generation process from a variety of elementary sources, is presented. A solution to the problem of detection of geoacoustic signal informative features is presented by the means of description of signal fragments by the matrixes of local extrema amplitude ratios and of interval ratios between them. The result of applying the developed algorithm to describe automatically the structure of the detected pulses and to form a pattern set is shown. The patterns characterize the features of geoacoustic emission signals observed at IKIR FEB RAS field stations. A technique of reduction of the detected pulse set dimensions is presented. It allows us to find patterns similar in structure. A solution to the problem of processing of a large data flow by unifying pulses description and their systematization is proposed. A method to identify a geoacoustic emission pulse model using sparse approximation schemes is suggested. An algorithmic solution of the problem of reducing the computational complexity of the matching pursuit method is described. It is to include an iterative refinement algorithm for the solution at each step in the method. The results of the research allowed the authors to create a tool to investigate the dynamic properties of geoacoustic emission signal in order to develop earthquake prediction detectors.

Analysis of geoacoustic emission and electromagnetic radiation signals accompanying earthquake with magnitude Mw = 7.5

“…And vice versa, if the number of local extrema continuously following one after another lies outside the specified interval, the detected signal fragment is not a pulse. Detailed description of the pulse detection procedure is presented in [6,7].…”

“…To identify and classify pulse waveforms, the structural description method is proposed. It is to encode a pulse with special binary matrix called "descriptive matrix" [7]. To construct the descriptive matrix, amplitudes of pulse local extrema and time intervals between them are compared:…”

A Clone Selection Algorithm Optimized Support Vector Machine for AETA Geoacoustic Anomaly Detection

“…For signal processing methods, in 2019, Senkevich et al [18] introduced a method to describe signal segments through local extreme amplitude ratios and extreme interval matrices to solve the problem of geoacoustic signal feature selection. In 2020, Lukovenkova et al [19] utilized signal preprocessing, an adaptive matching pursuit algorithm, and a pulse classification method based on description matrix similarity to process geoacoustic signals.…”

Search for anomalies in pulse flows of acoustic and electromagnetic emissions

“…The coefficient of symbolic overlap of the alphabets kAB for alphabets A and B with sizes N and M, respectively, is determined by the formula: Computer programs have been developed for methods for extracting pulses and composing alphabets of messages used to process and analyze emission geophysical signals. The sensitivity and noise immunity of the implemented algorithms were estimated using a numerical experiment [10]. We use the introduced definition of the alphabets intersection to assess changes in the pulse stream from one time episode of observation to another.…”