Geoacoustic inversion of the pressure gradient data using a single vector sensor

Abstract

A vector sensor has the capability of obtaining the acoustic field involving the sound pressure and three-dimensional particle velocities simultaneously, which means that there exists a potential to extract deep information for further applications like direction of arrival (DOA), positioning, classification, geoacoustic inversion, etc. This paper applies a nonlinear Bayesian approach to the pressure gradient data for geoacoustic properties of South China Sea, about which the experiment took place in July of 2013 in South China Sea of nearly 90 m depth. Recorded linear frequency modulation (LFM) signals of the pressure and vertical particle velocity are windowed and processed together to estimate the pressure gradient in the vertical direction at multiple frequencies as the observation data. After the optimal sediment model is selected based on BIC (Bayesian Information Criterion) using adaptive simplex simulated annealing (ASSA), delayed rejection adaptive Metropolis (DRAM), a Markov chain Monte Carlo (MCMC) sampling method, is used not only to provide a maximum a-posteriori (MAP) parameters estimates but also to quantify the parameters uncertainties and inter-parameter relationships. [Work supported by the National Natural Science Foundation (Grant No. 61531012).]