Auto clustering of the variety of pulse signals based on their symbolic description

Yury Senkevich,N Salikhov,A Vilayev,B Shevtsov

doi:10.1051/e3sconf/201912702005

Yury Senkevich, N Salikhov + Show 2 more

Open Access

https://doi.org/10.1051/e3sconf/201912702005

Copy DOI

Abstract

In a number of applied studies of geophysics, medicine, cosmophysics, atomic physics and other fields of knowledge, useful information is often hidden in the character of the behavior of a stream of frequency modulated pulses, which are represented by a large variety of forms, significantly different from each other up to several orders value of magnitude amplitude and durations. Noise is often present in the signal. Under these conditions, the problem arises of identifying both individual pulses and groups of pulses to assess the connection between their dynamic characteristics and the state of system. To solve the problem by a method is proposed that includes signal cleaning from interference, the operation of extracting and converting pulses into a code representing a sequence of invariant amplitude and time transformations of similar pulses combined by a single graphic pattern called “symbol”. All symbols extracted from the signal make up the alphabet. A procedure for narrowing the dimension of the alphabet is shown, which allows you to automatically divide it into clusters according to the degree of coincidence of the code. The results of the practical application of the developed method for the selection of base classes of the geoacoustic emission (GAS) signals related to the objective data of the state of the signal-generating medium are presented. The study used data from the archives of observations IKIR FEB RAS.

Highlights

Studies have shown that signals of geoacoustic emission indirectly reflect the characteristics of processes occurring in the lithosphere [1]
In the presence of sensors that allow fixing the accompanying physical fields, it is possible to take into account their influence on the radio signal
During the repetition of the experiment, it was found that the distribution obtained is resistant to changes in the initial conditions of randomization of the start of the generator of the pseudo-noise sequence of a computer program simulating Gauss noise

Summary

Introduction

Studies have shown that signals of geoacoustic emission (radio) indirectly reflect the characteristics of processes occurring in the lithosphere [1]. The physical nature of radio pulses is well studied and represents processes associated with changes in external fields [2,3,4], first of all, tense Earth's crust under the influence of tectonic movements, which contribute to the detection of seismic waves. The circumstances listed above make it difficult to obtain useful information in a detailed analysis of the radio Most often, such indicators as the integrated energy of incoming pulses are used for analysis [2] or dimension of the set of shapes of the emitted pulses [5]. The search of a connection to changes in the quantitative and qualitative characteristics of radio pulses during the manifestation of changes in the state of the lithosphere according to recorded seismic events is the purpose of the presented study

Description of the method and algorithms

Results of the nature experiments and its discussion

Findings

Research findings