2023_programme: Vertical Directionality of the Ambient Noise using Array Processing and PCRA.

• Session: 21. Underwater Noise - Modelling and Measurements
  Organiser(s): Alexander Gavrilov
• Lecture: Vertical Directionality of the Ambient Noise using Array Processing and PCRA.
  Paper ID: 1950
  Author(s): Real Gaultier, Berbey Nino, Pihan-Le Bars Hélène, Kinda Bazile
  Presenter: Real Gaultier
  Abstract: For underwater acoustic detection, knowledge of the characteristics of ambient noise is critical. In particular, for mid-frequency signals (above 1 kHz), the vertical directionality of ambient noise is of great interest. In presence of sound speed profile with a negative gradient, the sources of noise in this frequency band (surface agitation and local shipping) contribute considerably less to the ambient noise level. The resulting shape of the vertical noise directionality exhibits therefore a « notch », usually located around the horizontal direction.\nWhile previous studies show examples of observation of the noise-notch, they also demonstrate that relatively long period of stationary observations are required to evaluate its characteristics (depth and width). \nIn this study, we propose a processing scheme allowing : first to beamform the data in the time domain, second to remove the transient components of the ambient noise, using Percentile Controlled Recursive Averaging (PCRA) on each formed beam.\nPCRA is a Bayesian method which recursively evaluates clean signal and noise components from the estimation of a conditional signal presence probability in the time-frequency domain. In our case, it is used to remove the transient signals.\nThe method was evaluated on a dataset recorded during the ALMA2018 campaign. During this multi-site at-sea trial, an irregularly sampled planar array was deployed and recorded several hours long samples of ambient noise in various environments. For the sake of simplicity, the array was considered as a Vertical Line Array (VLA). Comparisons with full 2D frequency-domain beamforming without PCRA are also provided. \n\nThe results show the capability of the proposed method to filter out the transient signals and benefits from long stationary samples to highlight the presence of a noise notch.
• Corresponding author: Dr Gaultier Real
  Affiliation: DGA Naval Systems
  Country: France
  e-mail: