Non-parametric acoustic source depth discrimination in a ducted environment with a vertical line array

Underwater acoustic measurements made with vertical line arrays (VLAs) can exploit horizontal sound speed stratification to resolve dominant signal propagation paths and elevation angle-dependent ambient noise structure. Leveraging this attribute of vertical aperture, a non-parametric approach to source depth discrimination is introduced. The approach is distinguished by its lack of reliance on a priori knowledge of the sound speed environment or modeling of the channel's depth-dependent normal mode functions. Instead, the method relies solely on a measurement of the vertical wavenumber spectrum associated with the principal eigenvector of the array spatial cross-covariance estimate. It will be shown that, given sufficient vertical aperture, the principal eigenvector is an estimator of the dominant normal mode or modes excited by the source. Two methods for estimating the vertical wavenumber spectrum from the principal eigenvector are presented: conventional (or fast Fourier transform-based) and minimum variance distortionless response. For both methods, implementation issues and differences in spatial resolution will be examined. Depth discrimination performance will be quantified using receiver operating characteristic curves developed from at-sea experimental data collected on a 32-channel vertical line array deployed from a Liquid Robotics SV-3 wave glider (Liquid Robotics, Herndon, VA) in downward-refracting conditions off the coast of San Diego, CA, in August 2017.