Maximum entropy inference of seabed properties using waveguide invariant features from surface ships.

Acoustic data were recorded on two vertical line arrays (VLAs) deployed in the New England Mud Patch during the Seabed Characterization Experiment 2017 in about 75 m of water. The sound recorded during the passage of merchant ships permits identification of singular points for the waveguide invariant ß for mode pairs (1,n):ß1,n,for n=2,3,4,5, in the 15-80 Hz band. Using prior geophysical information and an acoustic data sample from the merchant ship KALAMATA, a geoacoustic model M of the seabed was developed. Then, using data samples from other merchant ships, a feature-ensemble maximum entropy method is employed to infer the statistical properties of geoacoustic parameter values for the sound speeds in a surface mud layer and a deep sand layer. Technical challenges include a sparsity of observed singular points, the unique identification of mode pairs for an observed singular point, and the deviation of the waveguide from horizontal stratification. A geoacoustic model M is developed that reproduced the observed ß≈-1 for f < 20 Hz and mode cutoff features at about 15 Hz. The statistical low-frequency inference of the singular point structure from multiple ships provides evidence of an angle of intromission at the water sediment interface with an average sound speed ratio of about 0.986 and an average sound speed for the deeper sand layer of about 1775 m/s.

Impact of data augmentation on supervised learning for a moving mid-frequency source.

Two residual networks are implemented to perform regression for the source localization and environment classification using a moving mid-frequency source, recorded during the Seabed Characterization Experiment in 2017. The first model implements only the classification for inferring the seabed type, and the second model uses regression to estimate the source localization parameters. The training is performed using synthetic data generated by the ORCA normal mode model. The architectures are tested on both the measured field and simulated data with variations in the sound speed profile and seabed mismatch. Additionally, nine data augmentation techniques are implemented to study their effect on the network predictions. The metrics used to quantify the network performance are the root mean square error for regression and accuracy for seabed classification. The models report consistent results for the source localization estimation and accuracy above 65% in the worst-case scenario for the seabed classification. From the data augmentation study, the results show that the more complex transformations, such as time warping, time masking, frequency masking, and a combination of these techniques, yield significant improvement of the results using both the simulated and measured data.

Learning location and seabed type from a moving mid-frequency source.

While source localization and seabed classification are often approached separately, the convolutional neural networks (CNNs) in this paper simultaneously predict seabed type, source depth and speed, and the closest point of approach. Different CNN architectures are applied to mid-frequency tonal levels from a moving source recorded on a 16-channel vertical line array (VLA). After training each CNN on synthetic data, a statistical representation of predictions on test cases is presented. The performance of a single regression-based CNN is compared to a multitask CNN in which regression is used for the source parameters and classification for the seabed type. The impact of water sound speed profile and seabed variations on the predictions is evaluated using simulated test cases. Environmental mismatch between the training and testing data has a negative impact on source depth estimates, while the remaining labels are estimated tolerably well but with a bias towards shorter ranges. Similar results are found for data measured on two VLAs during Seabed Characterization Experiment 2017. This work shows the superiority of multitask learning and the potential for using a CNN to localize an acoustic source and detect the surficial seabed properties from mid-frequency sounds.

Observations of scatter from surface reflectors with Doppler sensitive probe signals.

Previous analyses of surface scatter from the at-sea KAM11 experiment were made with linear frequency modulated waveforms that emphasized a single slope direction for arrivals in time-varying impulse response estimates. This analysis of Doppler sensitive waveform transmissions, made in the same geometry, resolves additional scatter arrivals with opposite slope. The different Doppler shifts in scatter observations are related to dispersed, naturally occurring, sea surface features that reflect the transmitted waveform to the receiver. The positions of these surface reflectors are estimated from the delay and Doppler shift of the observed arrivals without needing a receiving array with high spatial resolution.

Calibration of vertical array tilt using snapping shrimp sound.

Snapping shrimp are the dominant biological source of high-frequency (>2 kHz) ambient noise in warm coastal waters. In a recent experiment, the highly impulsive signals produced by shrimp snaps were recorded continually by a large-aperture vertical array (56 m) that was bottom-moored in 100-m deep shallow water. Assuming the array vertical, initial localization of individual snaps based on wavefront curvature along the array indicated that all snaps came from either above or beneath the flat seabed. By constraining all snaps to originate from the seabed, several hundred snaps within a radius of 500 m from the array over a 20-s window were detected successfully and localized in the three-dimensional space of time-of-arrival, range, and array tilt. Since the estimated array tilt for each snap is a projection of the absolute array tilt onto the nominal array-snap plane, the maximal tilt in the range and tilt domain corresponds to the absolute array tilt. Both simulations and data demonstrate that snapping shrimp can be exploited as a source of opportunity for calibration of vertical array tilt.

Acoustic scattering comparison of Kirchhoff approximation to Rayleigh-Fourier method for sinusoidal surface waves at low grazing angles.

The Fourier series method for implementing the Rayleigh hypothesis [Rayleigh-Fourier method (RFM)] is used as a reference solution to assess the Kirchhoff approximation of the Helmholtz integral [Helmholtz-Kirchhoff approximation (HKA)] for modeling broadband scatter from sinusoidal surfaces at low grazing angles. The HKA is a valuable solution because it has an eigen-ray interpretation without unbounded caustic amplitudes and discontinuous shadow zones. Plane wave studies of the HKA, however, show it becomes inaccurate at low grazing angles. This study quantifies how this limitation manifests with increasing transmission distance for time domain scattering simulations. Scattering results are compared over a complete surface wave cycle with parameters modeling sea surface-swell. The HKA agrees reasonably well with the RFM in point source calculations for limited extensions of transmission distances beyond where plane wave comparisons begin to diverge. Past these distances, HKA solutions begin to show significant over-prediction of the acoustic amplitude around late arrivals. This over-prediction is frequency dependent and eigen-ray interference offers an explanation of this behavior. Further extending the transmission range leads to a significant HKA error, and a range is found at which flat surface reflections have less error.

Passive acoustic tracking using a library of nearby sources of opportunity.

A method of localizing unknown acoustic sources using data derived replicas from ships of opportunity has been reported previously by Verlinden, Sarkar, Hodgkiss, Kuperman, and Sabra [J. Acoust. Soc. Am, 138(1), EL54-EL59 (2015)]. The method is similar to traditional matched field processing, but differs in that data-derived measured replicas are used in place of modeled replicas and, in order to account for differing source spectra between library and target vessels, cross-correlation functions are compared instead of comparing acoustic signals directly. The method is capable of localizing sources in positions where data derived replicas are available, such as locations previously transited by ships tracked using the Automatic Identification System, but is limited by the sparsity of ships of opportunity. This paper presents an extension of this localization method to regions where data derived replicas are not available by extrapolating the measured cross-correlation function replicas onto a larger search grid using waveguide invariant theory. This new augmentation provides a method for continuous tracking.

Toward ocean attenuation tomography: Determining acoustic volume attenuation coefficients in seawater using eigenray amplitudes.

A deep-water experiment in the Pacific made in situ measurements of the volume attenuation coefficients of sea water in the mid-frequency range. The frequency, temperature, salinity, pH, and pressure dependent seawater attenuation coefficients were determined using a vertical line array that received and identified over 2000 unique paths from 1200 mid-frequency 3-9 kHz LFM source transmissions at a convergence zone range and depth up to 400 m. The results show no change in attenuation coefficients in this band compared to estimates from 30-year-old models previously determined from a combination of long-range ocean acoustic and laboratory experiments. The inversion also explores the feasibility of ocean acoustic attenuation tomography to further separate the depth-dependent chemical components responsible for the total attenuation loss through by isolating a group of deep-water refracting acoustic paths.

Robust source-range estimation using the array/waveguide invariant and a vertical array.

The array invariant based on beam-time migration has been proposed for source-range estimation using a horizontal or vertical array (VA), with no need for forward model computations. The approach has been demonstrated successfully with experimental data in shallow water using a horizontal towed array. Recently, the array invariant has been shown to be a special case of the waveguide invariant theory. In this paper, the unified array/waveguide invariant approach to source-range estimation is applied to a short-aperture VA in a fluctuating ocean environment over a one-day period. Specifically, the mean range estimates using a 12-m long VA in ∼100 m deep water are <8% relative error for a source (2-3 kHz) at 6-km range, demonstrating the robustness of this approach.

Passive acoustic source localization using sources of opportunity.

The feasibility of using data derived replicas from ships of opportunity for implementing matched field processing is demonstrated. The Automatic Identification System (AIS) is used to provide the library coordinates for the replica library and a correlation based processing procedure is used to overcome the impediment that the replica library is constructed from sources with different spectra and will further be used to locate another source with its own unique spectral structure. The method is illustrated with simulation and then verified using acoustic data from a 2009 experiment for which AIS information was retrieved from the United States Coast Guard Navigation Center Nationwide AIS database.

Self-synchronization and spatial diversity of passive time reversal communication.

Time reversal (TR) achieves spatial and temporal focusing in complex environments. In the context of communications, passive (uplink) TR is equivalent to active (downlink) TR with the communications link being in opposite directions but with the same theoretical performance. The benefit of passive TR, however, is its natural self-synchronization, making TR communications robust and readily extended to cases involving temporal diversity and synthetic aperture communications. Self-synchronization is examined analytically and demonstrated with shallow water experimental data. In addition, the impact of spatial diversity on communications performance is investigated in terms of the aperture and element spacing of the 64-element vertical receive array.

Efficient use of bandwidth for underwater acoustic communication.

In a recent shallow water experiment, acoustic communication transmissions were carried out over the 10 to 32 kHz band in ~100 m deep water over a 3 km range. A natural question is how best to utilize that bandwidth. In one multiband approach discussed previously, the band was divided into four smaller subbands that were processed independently using time reversal decision-feedback equalizers (TR-DFEs). This letter presents a complementary wideband approach using data from the same experiment achieving a data rate of up to 60 kbits/s with 32 quadrature amplitude modulation. These results suggest that a wideband approach can be beneficial in terms of spectral efficiency with modest computational complexity using a TR-DFE.

Analyzing sound speed fluctuations in shallow water from group-velocity versus phase-velocity data representation.

Data collected over more than eight consecutive hours between two source-receiver arrays in a shallow water environment are analyzed through the physics of the waveguide invariant. In particular, the use of vertical arrays on both the source and receiver sides provides source and receiver angles in addition to travel-times associated with a set of eigenray paths in the waveguide. From the travel-times and the source-receiver angles, the eigenrays are projected into a group-velocity versus phase-velocity (Vg-Vp) plot for each acquisition. The time evolution of the Vg-Vp representation over the 8.5-h long experiment is discussed. Group speed fluctuations observed for a set of eigenrays with turning points at different depths in the water column are compared to the Brunt-Väisälä frequency.

Multiuser acoustic communications with mobile users.

A multiuser receiver is developed that is capable of separating receptions from independent, mobile users whose transmissions overlap in both time and frequency. With respect to any one user's Doppler corrected signal, the other communication signals appear as multiple-access interference distributed across the Doppler dimension. A previously developed receiver composed of an adaptive time-reversal processor embedded within a successive interference cancellation framework is limited to stationary users. This paper extends the receiver to properly remove the interference from moving sources by modeling the effects of Doppler through the interference cancellation receiver. The combined receiver has the ability to remove interference in both the temporal and spatial domains, and this property is shown to be preserved even when users are in motion. When applied to data collected during a recent shallow water experiment (KAM11), the receiver is shown to be capable of separating packets in a two user system where one user is moving while the other is stationary.

Diversity combining for long-range acoustic communication in deep water.

A recent experiment showed that coherent long-range acoustic communication is feasible in deep water over a ∼550 km range between a source towed slowly at ∼75 m depth and a horizontal line array towed at 3.5 knots at ∼200 m depth. This letter further demonstrates that diversity combining mitigates channel fading and increases the output signal-to-noise ratio. Using sparse channel-estimate-based equalization, three transmissions are combined successfully to decode a 40 Hz bandwidth (230-270 Hz) 8 phase-shift-keying communication signal, achieving an effective data rate of 17 bits/s at ∼550 km range.

Asynchronous multiuser underwater acoustic communications (L).

An asynchronous multiuser system is proposed to support multiple-access underwater communications without the use of code-division multiple-access or a feedback channel. The rich multipath channels experienced by spatially separated users will be sufficient to ensure separation of collided packets at the base station. The iterative receiver will employ a combination of adaptive time-reversal processing, matching pursuit, and successive interference cancellation in a block-wise fashion to achieve multiuser separability. Data collected during the KAM11 experiment are used to illustrate the system's capability in a dynamic, time-varying environment.

Long-range multi-carrier acoustic communication in deep water using a towed horizontal array.

During a recent long-range acoustic communication experiment carried out in deep water, multi-carrier Orthogonal Frequency Division Multiplexing (OFDM) communication signals were transmitted with a 50 Hz bandwidth (225-275 Hz) at various source-receiver ranges from 100 to 700 km. The experiment consisted of two mobile components: (1) a source towed slowly at a speed of 2-3 knots at ∼75 m depth and (2) a horizontal line array towed at 3.5 knots at a depth of ∼200 m. In addition to beamforming, an interleaver gain is exploited to compensate for low signal-to-noise ratio at the expense of data rate while providing diversity in the frequency domain. Error-free performance is shown at effective data rates of 15 and 7.5 bits/s at ranges of 550 km and 700 km, respectively, by combining interleaved repetitions with low-density parity-check coding after beamforming, demonstrating the feasibility of multi-carrier OFDM communications in deep water using a towed horizontal array.

Multiuser interference cancellation in time-varying channels.

In this letter, an adaptive time-reversal multichannel combiner is embedded within an iterative successive interference cancellation receiver. With the addition of matching pursuit, a sparse channel estimation technique, the combined receiver is shown to provide both temporal interference cancellation as well as spatial interference suppression in decoding simultaneous transmissions from separate users in a time-varying underwater acoustic environment. Experimental data collected during the KAM11 experiment illustrates that for a two-user multiple-access system, multiuser separation can be achieved.

Stochastic channel simulator based on local scattering functions.

This letter addresses the limitations of the stochastic channel simulation approach employed in wireless channels when directly applied to underwater channels. First it is shown analytically why the simulation method fails when the correlated taps each have a different Doppler spectrum. Then, based on the idea of local scattering functions, we propose a simple solution to the problem of simulating a channel with correlated taps where the correlated taps occur in subgroups with each of the subgroups having a different Doppler spectrum, applicable to underwater channels. Our simulation approach is tested successfully on the KAM08 channel.

Long-range acoustic communication in deep water using a towed array.

In September 2010 a long-range acoustic communication (LRAC10) experiment was carried out in deep water off the Southern California Coast. The experiment involved two mobile components: (1) a source towed slowly at a speed of 2-3 knots at ∼75-m depth and (2) a horizontal line array towed at 3.5 knots at a depth of ∼200 m. Phase-coherent communication sequences were transmitted in the frequency band of 200-300 Hz at various ranges (100-700 km). Initial analysis of the LRAC10 data demonstrates that an information rate of 50 bits/s can be achieved over ∼550-km range using quadrature-phase shift-keying (QPSK) modulation and error-correction coding combined with beamforming.