Neural networks for the localization of biological and anthropogenic source at neutrino deep sea telescope

Abstract

Astrophysics and marine biology come to work closely at underwater observatories which can be used on the one hand for the light and acoustic detection and localization of high energy neutrinos and on the other hand to acquire information about deep sea marine organisms and in particular marine mammals. This article describes some of the techniques developed for the acoustic localization of both anthropogenic (i.e. impulsive and tonal noises from shipping) and biological sources (i.e. marine mammals) in the vicinity of ANTARES. An overview of the methods and results for obtaining both the real-time locations of a set of 36 moving and widely spread acoustic sensors and the locations of nearby acoustic sources is given Besides the use of a matched filter for time-delay estimation and of steered beamformers, the main novelty of this work lies in the use of neural networks for localization which is of great interest for problems which are normally not tractable within a real-time frame. This solution is consequently a suitable approach for real-time passive acoustic monitoring when using the growing range of systems requiring low energy consumption and thorough resource management such underwater vehicles and wavegliders.