Gordon R. Ebbeson

A matched-field backpropagation algorithm for source localization

Model-based signal processing techniques have been developed over the years to improve the capability of active and passive sonar systems for detecting and localizing quiet underwater targets. In a generic matched-field processor, hydrophone signals measured at the array are compared to hypothetical signals (replicas) that are calculated by a full-field acoustic model for a given target position. This matching is carried out for many potential target locations within a search region (range, depth and bearing) to form an ambiguity surface whose peak values provide the greatest likelihood that targets are present. In this paper, we evaluate a version of a matched-field processor that combines measured data with a higher-order parabolic equation (PE) algorithm to effectively backpropagate an (unnormalized) ambiguity surface outwards from the receiving array. To illustrate this PE-based method, the unconventional processor is applied to some synthetic and experimental hydrophone data received on vertical line arrays in shallow-water waveguides.

A parabolic equation‐based backpropagation algorithm for matched‐field processing

In applications of matched‐field processing (MFP) to passive sonar, signals measured on a hydrophone array due to a source at some unknown position xs=(rs,zs) are correlated with signals predicted by a propagation model (replicas) due to a source at a given position x=(r,z). This matching is carried out for many assumed x’s within a search region (range, r and depth, z) to form a (normalized) ambiguity surface whose peak value provides an estimate of xs. For an N‐element array, reciprocity is invoked to reduce the computational effort to computing N replica fields at each point on the search grid. In this paper, a matched‐field processor proposed by Tappert et al. [J. Acoust. Soc. Am. 78, S85 (1985)] is revisited that combines measured data with PE starting fields to effectively backpropagate an (unnormalized) ambiguity surface outwards from the receiving array. This unnormalized processor generates the ambiguity surface N times faster than the normalized one. The effectiveness of the backpropagation algo...

A generalized beamformer for tracking tonal sources

A generic matched-field processor is used to track tonal sources recorded on an L-shape array deployed on the seafloor in St. Margarets Bay, Nova Scotia, Canada. The replicas are computed using a normal mode summation based on the Zhang and Tindle model for isovelocity water over an attenuating fluid bottom. The model uses an effective depth approximation to solve for the modes, which makes the code computationally efficient. Localization accuracy for frequencies in the 100- to 300-Hz range was estimated for this environment using known signals from a towed source

Acoustic propagation in an Arctic shallow-water environment

A short-range, shallow-water transmission loss experiment was conducted in the South Lincoln Sea, under an irregular rough ice canopy. A barrel-stave source was used to transmit 800-Hz tonal signals at depths of 10 and 30 m. The signals were received on a four-element vertical line array distributed over the 65-m water depth. To support the transmission loss measurements, a bottom loss experiment was carried out using broadband signals generated by imploding light bulbs at three depths and with data received on the same hydrophone array. Previous drilling experiments in the vicinity of this site have provided seabed information and revealed the presence of a subbottom permafrost layer. Preliminary modeling of the acoustic propagation in this highly irregular channel was carried out using a wave-based model. This paper reviews the measurements, and discusses the challenges associated with the modeling, in view of the experimental uncertainties associated with this complicated environment.

A PE-BASED APPROACH TO MODAL DECOMPOSITION AND SOURCE LOCALIZATION

Matched mode processing is an efficient alternative to matched field processing for locating a source in an acoustic waveguide. The successful application of this method relies on the accurate modal decomposition of the pressure data received on a vertical line array. From the modes that can be resolved by a given array aperture, the source range is determined from modal phases, whereas the source depth is determined from the mode shapes. In this paper, we describe a decomposition method based on the parabolic equation (PE) propagation model for recovering this modal information from vertical array data and which can be used to localize an underwater source by the matched mode processing method. We apply our PE-based decomposition/localization scheme to several benchmark problems involving a narrowband acoustic source operating in a shallow water environment.

Broadband source localization using matched correlation processing

For many years, model‐based signal processing algorithms using Matched Field Processing (MFP) techniques have been analyzed with the goal of improving the capability of passive sonar systems for localizing quiet underwater sources. Recently, researchers at DRDC Atlantic have been investigating Matched Correlation Processing (MCP) as a faster alternative to MFP. In this method, the cross‐correlations for a source as measured with a pair of hydrophones in a horizontal array are matched with those generated with a correlation model for many candidate ranges and depths along a candidate bearing. These matches are carried out with a number of hydrophone pairs to form many ambiguity surfaces. The maximum on the average of these surfaces is assumed to yield the best estimate of the source position. By carrying out this procedure over a number of candidate bearings, a full 3‐D search for the source location is achieved. Since 2002, a number of localization trials have been carried out east of Nova Scotia, Canada....

Matched‐field geoacoustic inversion in a range‐varying waveguide

In recent years, geoacoustic inversion methods based on matched‐field processing concepts have been devised for inferring the sub‐bottom properties of shallow‐water waveguides using measurements of acoustic signals received on hydrophone arrays. The development of these matched‐field inversion (MFI) procedures is driven by two goals: (1) to obtain an accurate geophysical description of the morphology of the sub‐bottom structure, and (2) to characterize the acoustic response of the sub‐bottom so that its effect on sound propagation can be predicted. Knowledge of (2) can improve the effective ranges for using sound in underwater communication and/or source localization applications. Of practical interest is the question: How crude can our estimates of the sub‐bottom structure be and still allow successful acoustic source localization using model‐based signal processing methods? In this paper, a hybrid simplex simulated annealing MFI code [M. R. Fallat and S. E. Dosso, J. Acoust. Soc. Am. 105, 3219–3230 (199...

PE solutions to some internal‐wave benchmark problems

Benchmarking underwater acoustic propagation models is a necessary stage in the evolution of numerical modeling codes. Candidate benchmark problems should challenge the capabilities of existing models in order to promote the development of improved techniques for reliably simulating sound propagation in realistic oceans. In this paper, a parabolic equation (PE) model is applied to the suite of internal‐wave test cases that are offered for numerical consideration by the organizers of the Benchmarking Range Dependent Numerical Models session. The PE calculations are carried out using a code that was originally developed for matched‐field processing applications [G. H. Brooke et al., ‘‘PECan: A Canadian parabolic equation model for underwater sound propagation,’’ J. Comput. Acoust. (2001)]. The capability of PECan to propagate sound accurately through the benchmark environments is examined for both tonal and broadband signals. The discussion will address issues related to reciprocity, interpolation of the in...

PE modal decomposition of VLA data

The first step in any matched‐mode processing application is the decomposition of the measured acoustic field into its horizontal wave‐number (modal) components. Previously, a PE‐based method was described for carrying out the modal decomposition of the pressure field received on a vertical line array (VLA) [D. J. Thomson et al., J. Acoust. Soc. Am. 89, 2000 (1991)]. For that method, the VLA field was first interpolated between hydrophones and then extrapolated into the ocean bottom. This interpolation/extrapolation stage was needed to construct a suitable starting field on the fine computational grid used by the PE algorithm. In this paper, a modified PE model is presented that eliminates the computational grid in the bottom by applying an exact, nonlocal boundary condition along the ocean‐bottom interface. Consequently, it is no longer necessary to extrapolate the VLA field prior to propagating it with PE. To demonstrate the effectiveness of the boundary condition, PE‐based modal decomposition is carrie...

Modal decomposition of the pressure field on a vertical line array

For many applications, it is useful to decompose the acoustic field measured in shallow water into its horizontal wave‐number components. Recently, a PE‐based method was described for effecting the modal decomposition of the depth‐dependent field at a given range in a range‐dependent waveguide [D. J. Thomson, ]. Acoust. Soc. Am. Suppl. 1 86, S53 (1989)]. This method should be applicable to the analysis of data obtained with a vertical line array (VLA). However, for practical arrays, the measured field is known at only a limited number of hydrophones, whereas the PE‐based decomposition method requires the field to be known at each depth on the computational grid. Therefore, to populate the entire PE grid, it is necessary to interpolate the field between hydrophones and to extrapolate it into the bottom. An interpolation and extrapolation scheme suitable for this purpose is proposed. To illustrate the effectiveness of this reconstruction scheme, modal decomposition is carried out using simulated VLA data ge...