Options
Wavelet decomposition and advanced denoising techniquesn for analysis and classification of seismic signals
Editor(s)
Language
English
Obiettivo Specifico
1.1. TTC - Monitoraggio sismico del territorio nazionale
Status
Published
Pages Number
118-129
Refereed
Yes
Title of the book
Issued date
2009
ISBN
978-1-905254-39-2
Abstract
This work describes an automatic classification procedure for seismic signals
suitable for the analysis of complex, broad-band waveforms commonly
associated with fluid-rock interaction in volcanic and hydrothermal systems.
Based on Discrete Wavelet Transform, a set of significant seismic signal
features that characterize the type of event is identified (e.g. noise, volcano
tectonic, long period). These features are initially assessed for events whose
category (class) can be previously determined by an expert analyst. A Bayesian
Pattern Recognition supervised technique based on these features is adopted
to classify a new ‘unlabelled pattern’, whose class is unknown. In this way
values computed for known events are used to classify events of unknown
identity ('supervised classification'). A test was performed on seismological data
recorded at Campi Flegrei (Italy), which was divided into three classes.
Automatic classification accuracy ranges from 82% to 100% over a broad range
of datasets.
suitable for the analysis of complex, broad-band waveforms commonly
associated with fluid-rock interaction in volcanic and hydrothermal systems.
Based on Discrete Wavelet Transform, a set of significant seismic signal
features that characterize the type of event is identified (e.g. noise, volcano
tectonic, long period). These features are initially assessed for events whose
category (class) can be previously determined by an expert analyst. A Bayesian
Pattern Recognition supervised technique based on these features is adopted
to classify a new ‘unlabelled pattern’, whose class is unknown. In this way
values computed for known events are used to classify events of unknown
identity ('supervised classification'). A test was performed on seismological data
recorded at Campi Flegrei (Italy), which was divided into three classes.
Automatic classification accuracy ranges from 82% to 100% over a broad range
of datasets.
Type
book chapter
File(s)
Loading...
Name
VOLUME_Galli et al_2009.pdf
Size
395.11 KB
Format
Adobe PDF
Checksum (MD5)
031818c21c385601fa3743149e9d9a77