Prof. Dr. Antonio Rubio
University of Granada
Department of Electronics & Computer Science

Automatic Identification of Volcanic Seismic Events

After explaining the context in which the Research Group of Speech Technology at the University of Granada is facing this new research adventure, the problem of classification of volcanic seismic events is introduced.

The physical origin of the volcanic signals is somehow similar to the origing of vocalic signals. Therefore, it could be expected that methods applied to speech recognition would apply also to idientifying different types of volcanic seismic events.

In addition, experts recognize very few types of events, which is a good situation for our purpose. The problem here is that we can not isolate signals removing noise, i.e, no clean signal can be used to train the system.

Nevertheless, we found that using the Hidden Markov Models paradigm we can identify correctly up to 86% of events, We are not using any kind of "event grammar" yet, but we expect that this could be useful in the future.