On the measurement and the modelling of the correlated noise sources corrupting the marine geological echosounding experiments

The modeling of the power spectral density of correlated noise sources in acoustic measurements in a marine environment is considered. The Belgian oceanographic vessel, the Belgica, was used for measurements of the noise generated by helixes, engines, stabilizers, the 'fish' incorporating the transducers, the waves, and the marine background noise. Classical estimators, such as the periodogram, as well as modern spectral estimators based on ARMA (autoregressive moving average) and AR (autoregressive) approaches are considered. It is shown that low-order parametric models are adequate to describe the observed and measured correlated noise sources. These models have made it possible to validate the synthetic echograms which are required for the development of algorithms solving the inverse problem.<>