Resumen:
The Ring of Fire represents a constant threat in the case of possible eruptions due to its 160 active volcanoes, in such way, several studies have determined that the increment of volcano tectonic and long period events could indicate the beginning of a volcanic eruptive process. In this context, it is crucial to improve the performance in microseisms detection, due to the high requirements towards automatic recognition systems in real-time. For this purpose, this work aims at improving a microseisms detector based on voice activity detection (VAD) and endpoint detection. Furthermore, we propose the use of signal segmentation, to work with fixed windows, to analyze the signal-to-noise ratio, and to adapting VAD techniques to microseisms. The results obtained demonstrate the reliability of this detector, by obtaining 100% of precision and 96% of accuracy in real-time tested, by considering dataset of Cotopaxi Volcano in Ecuador.
