Search

The description of underwater soundscape is central to the understanding of the marine environment, both from the standpoint of the fauna and anthropic activities and its interactions with the atmosphere. Some of these sources produce signals whose patterns are periodically repeated over time (i.e., ship propellers in motion, odontocetes clicks, snapping shrimp, noise emanating from surface waves, etc.). As ocean noise is a combination of various sources sometimes sharing the same frequency band, it is necessary to develop efficient algorithms to process the increasingly voluminous data acquired. To this end, the theory of cyclostationarity is adopted as an effective tool for exposing hidden periodicities in low signal to noise ratio. This theory, widely used to analyze mechanical systems or communications, is extended and applied on underwater soundscapes. The method is demonstrated using data recorded in the Celtic Sea at the French coast of Brittany with practical experiments using field measurements obtained from recording stations.

International audience; In an underwater acoustical context, our application concerns a system immersed, with Nt transmitters and Nr slowly moving receivers. The objective is that all receivers detect the transmitted signals, to estimate the time of arrival (TOA) and then to make possible the location with several TOA (more than 3). To use the CDMA (Code Division Multiple Access) modulation, specially adapted to our problem, we have to choose a method to generate a number Ns of broad-band signals. This study is devoted to the selection of Nt signals among the Ns. The aim is to choose the most distinctly detectable ones. Firstly, in a no Doppler context, the criterion of signals selection is based on a ratio between maximum of auto-correlation and cross-correlation. Secondly, in presence of Doppler, we rely on Ambiguity Function which allows representing the correlation function to several frequency Doppler shifts. The choice of Nt signals is then based on ratio between maximum of auto-ambiguity and cross-ambiguity. This paper will then highlight the interest of the criteria (correlation, ambiguity function) in the choice of the most appropriate signals in function of the multi-users context.