Symbol approximation in the problem of detection and analysis of precursor anomalies in geochemical monitoring data

• download

In this paper we consider an approach to segmentation of slowly changing telemetric signals on the base of
the method of symbol approximation Symbol approximation assumes the replacement of an initial signal with
sequence of symbols. Each symbol corresponds to uniquely determined local behavior of a signal. The behavior
may be specified in various ways and correspond to various local models. The most simple and defensible is the
set of symbols corresponding to time forms of a signal (increase, decrease, constant value). The further analysis
is directed to the search of behavior typical forms (patterns) among symbol sequence. Symbol approximation
method allows us to decrease the space of the considered dimension that decreases the algorithm computational
complexity, moreover the results obtained by this method are simple to interpret. In the paper the technique
of typical and abnormal forms of signal behavior in temporary realm and an analysis method for precursor
anomalies by the means of this method of segmentation are offered. The approaches described in the report
were used in the processing of geochemical monitoring data of Kamchatka.