Software C.I.E.E.V. | EGI Applications Database

Applications Database
Supporting
Applications Database
Supporting
Applications Database
Supporting
Applications Database
Supporting
Applications Database
Supporting

Click to visit the AppDB VMOps Dashboard for deploying and managing Virtual Machines to the EGI Cloud infrastructure.read more

.noscript.softwareentry {display: block;width: 1000px;}.noscript.softwareentry .field {display: block;padding-bottom: 0.5em;padding-left: 10em;width: auto;text-align:left;}.noscript.softwareentry .field .fieldtype {color: #444444;display: inline-block;font-weight: bold;min-width: 90px;vertical-align: top;width: 90px;}.noscript.softwareentry .field .fieldvalue {display: inline-block;max-width: 780px;text-align:left;}.noscript.softwareentry .field.image {height: 100px;left: -110px;position: absolute;top: 0;width: 100px;.height: 100px;border:none;}

Name:C.I.E.E.V.

Description:Clustering of infrasound events on Mt. Etna volcano

Abstract:The goal of this application is to make clustering of explosion events automatically, recorded by infrasound sensors, placed on Mt. Etna volcano. The application is composed by of five modules: Events triggerin system Cross-correlation clustering system Features extractor system Physical description Neural network supervised classification system Triggering system We get the events by using a method based on standard deviation squared (S.D.S) series of the infrasonic signal. We use two thresholds on the S.D.S series and we declare the events by using the upper one and their onset and ends points by using the lower one. Cross-correlation clustering system We classify the events using a cross correlation algorithm (Green and Nueberg 2006), where a maximum correlation coefficient n X n matrix, called M, for a buffer containing n events is built. The element mxy represents the maximum of the correlation function between the two events x and y, changing the position of the event x with respect to the event y, with a specific overlap, until the maximum is found. We adopt a modified version of the algorithm using another two thresholds in this step. Features extractor system From every family found, we get two features that give us the idea of the performance of the classification and to make more selective the clustering done at previous step, trying to decrease the number of the families found. The two features are, frequency peaks and the Q factor obtained by using the sompi method. Physical description this step is referred to the localization of the explosions with the semblance method, and to the time distribution of the events. This step gives physical information about the families found. Neural network supervised classification system Finally we train a neural network to classify events never seen, into the families found, using the algorithm described above, automatically. The algor