Source depth estimation based on Gaussian processes using a deep vertical line array

Abstract

For a bottom-moored vertical line array in the direct arrival zone, interference patterns have been used for source depth estimation. The interference pattern shows periodic modulation. Its period is directly related to the source depth, source frequency, and grazing angle. The performance degrades when the interference pattern is corrupted by ambient noise and other interferers. In this paper, broadband interference fringes are modeled as Gaussian processes (GPs) with a periodic kernel and are denoised using Gaussian process regression. The source depth is estimated based on the periodicity of the denoised interference fringe. Simulation results demonstrate that compared to the Fourier transform-based method, GPs provide a better performance with a low signal-to-noise ratio and a better ability to estimate the depth of a very shallow source. Real data recorded by a 105 m-aperture vertical array also verify the performance of GPs on source depth estimation without knowing the ocean environment.