SeGaMas : Serious Game for Marine Mammals Survey

Studying cetaceans is a complex task, due to the inaccessibility of the marine environment and low visibility. Passive acoustics is a very promising non-invasive solution. Nowadays, detection and classification of recorded species are carried out automatically and with good accuracy, in particular using AI methods. However, using such methods to locate animals is still imprecise, due to the lack of ground truth data on their actual movements.

To overcome this, we are working on the creation of a complete model of acoustic scenes, called SeGaMas (Serious Game for Marine (mammals) Survey). This model includes the generation of realistic cetacean trajectories, inspired by [1] and [2] , the regular emission of a biophonic signal, and a ray-tracing model to reconstruct the signal arriving at the sensor. As inputs, the model receives bathymetric, oceanographic and ambient noise data. As output, it provides animal trajectories, the received signal and the paths taken by this signal.

These simulated data are then used to create a “click sequences to trajectories” AI model, capable of reconstructing the trajectory of a sound source from a simple recording. They will also be useful in the study of cetacean perceptions, and to better design and position our sensors.

[1] Parrott et al (2010). 3MTSim: An agent-based model of marine mammals and maritime traffic to assist management of human activities in the Saint Lawrence Estuary, Canada.

[2] Chouchane et al (2012), EvoCOP 2012: Splitting method for spatio-temporal sensors deployment in underwater systems