Zaugg, S., van der Schaar, M., Riccobene, G., André, M.
Real time detection and classification of natural biological and anthropogenic sounds from underwater antennas
Proceedings of the 23rd Annual Conference of the European Cetacean Society, Istanbul, Turkey, p.66, Mar 2009

Abstract:
Passive acoustic monitoring (PAM) is a powerful tool for the study of marine mammals and anthropogenic sounds in the marine environment. However PAM campaigns result in huge quantities of data. In this context it is desirable to reduce the data volume in order to alleviate the burden put on data transmission, storage and analysis. To address this problem we developed a set of algorithms that work in real time and in a fully automated mode. The algorithms perform two tasks: (1) They detect acoustic events (e.g. cetacean calls, ship noise, explosions), thus making possible the automated discarding of data segments that contain only sea noise. (2) They tag the remaining data into broad groups (e.g. impulsive sounds, constant tonals, frequency modulated tonals). This allows to pass only specific subsets of the data to more sophisticated and time consuming analyses. The performance of our algorithms was assessed on datasets from several deep see platforms. These datasets encompass all the diversity of sounds that must be expected in long term PAM campaigns (e.g. they also contain data segments with low SNR or with the simultaneous presence of different kind of sounds). The achieved performance indicates that the algorithms are very reliable and hence can be used as a data processing step in deep sea PAM campaigns. During the presentation a RT detection and classification of recorded data will be performed and displayed.

Project: LIDO, Listening to the Deep-Ocean Environment

Project: ESONET, Network of Excellence