John M. Ozard

Matched field processing in shallow water for range, depth, and bearing determination: Results of experiment and simulation

Recent theoretical studies have investigated high‐resolution methods for estimating source position in a shallow‐water waveguide. In this paper, an eigenvector method and a geophysical model based on several types of seismic data are used to estimate the bearing, depth, and range of a moving submerged source in the presence of an interfering surface source at the same range and bearing. The receiver was a sparse horizontal array of four sensors on the bottom (depth 82 m). A variant of the eigenvector method known as the MUSIC method was employed for measured and test field matching and a normal mode model was employed to calculate the test field. The data analyzed were obtained by towing a submerged source directly below the stern of the CFAV Endeavour, so that the surface and submerged sources were essentially at the same range and bearing. Source bearing estimates for the data showed that source bearings were usually well defined although somewhat ambiguous. The presence of the deep source was indicated...

Matched‐field processing in a range‐dependent environment

Matched‐field processing (MFP) is being considered increasingly for three‐dimensional (3‐D) localization of an acoustic source in a noisy ocean environment. MFP consists of comparing the measured acoustic field to the full field computed using a geophysical and a propagation model. Most MFP implementations have involved only range‐independent propagation models, and many have been restricted to vertical arrays. However, many realistic environments cannot be adequately described by range‐independent models, and the problem of localization using more general arrays is of increasing interest. In this paper, a technique is described for range‐dependent MFP with arbitrary arrays, where the field is computed using a parabolic equation (PE) approximation. Using PE, two‐dimensional (2‐D) field values are computed for each sensor in the array for a set of possible source ranges, depths, and (N) bearings to form an N×2‐D field model. Discrete estimates of the position of the source are obtained by applying MFP, wit...

An artificial neural network for range and depth discrimination in matched field processing

Associative feedforward neural networks with no hidden layers were applied to the problem of localizing a source in range and depth using the acoustic signal arriving at a vertical array of sensors. A highly processed form of the signal (excitations of an orthogonal basis) was used as input in order to increase the robustness of the trained network. The output layer consisted of one unit for each possible range and one unit for each possible depth of the source. The networks were trained with a signal‐to‐noise ratio (S/N) at the hydrophone of 50 dB, and then their performance was evaluated with S/Ns of 50 and 0 dB. Network weights were found for narrow and broad target shapes that correspond to narrow and broad beamshapes. The narrow target produced the beamformer with the lowest sidelobes and highest gains with acceptable but somewhat higher sensitivity to noise. Performance in the region for which the network was trained compared favorably with minimum variance beamforming.

Matched‐field processing using a neural network with preprocessing

Model‐based signal processing for source localization usually requires a comparison of the measured and replica acoustic fields at all possible source positions. It is anticipated that this problem may be solved more efficiently as a pattern recognition problem. In order to make such a solution applicable to arbitrary array shapes, the covariance matrix of the simulated data was preprocessed and represented as the excitations of the eigenvectors. The orthogonal bases for the total possible signal space was employed. Localization of a source in depth was performed using a linear perceptron with the excitations as input. The perceptron was trained with the excitations for a distribution of source depths. Precision of localization of the resulting processor and its robustness in the presence of noise were measured and compared to the performance of a minimum variance matched‐field processor. Extensions of the processor to estimate range and azimuth are being investigated by using a multilayer neural network ...

Artificial neural networks for simultaneous and independent range and depth discrimination in passive acoustic localization.

Two feedforward neural networks (NNs) with one hidden layer each were trained using a modified backpropagation algorithm to determine the position of an acoustic source in a waveguide. One network was trained to localize the source in depth while the other was trained independently to localize in range. The signal was preprocessed by decomposition along an orthogonal basis vector set in order to increase the robustness of the resulting trained network to uncertainties in the signal and environmental parameters. The output layer consisted of one unit for each possible range or depth of the source. The NNs were trained with a signal‐to‐noise ratio (S/N) of 50 dB and tested with patterns generated with S/Ns ranging from 50–0 dB. The performance of the NNs was compared with that of a conventional nearest neighbor processor. Evaluation of the processors was done in the context of an estimation problem, i.e., by measuring the standard deviation of the processors’ outputs. The NNs turned out to be less resistant...

Improving performance for matched‐field processing

Minimum energy beamforming (MEB) is frequently employed in matched‐field processing (MFP) because it is optimal for the noise field prevailing at the time of measurement. However, when signal phase errors or mismatch are present, MEB is no longer optimal for signal detection. Under such conditions various methods such as reduced minimum energy beamforming (RMEB), modal beamforming, and dimension reduction have been employed. For vertical arrays, these methods have been shown to be virtually identical and to improve performance dramatically over that obtained with MEB. In this paper RMEB is extended to arbitrary array configurations and the normalizations of the ambiguity surface for various noise fields are investigated. The generalized form of RMEB is shown to produce significantly improved performance for both vertical and horizontal arrays and thus demonstrates the benefits of modal style beamforming for horizontal arrays. The improved performance was obtained for both modal and isotropic noise fields....

Matched‐field processing and the effect of realistic noise correlation models

In simulating acoustic source localization by matched‐field processing, it is desirable to use a realistic ambient noise model. Two general classes of noise model were considered: a planar isotropic noise source distribution and a line source distribution. The former was implemented based on the work of Kuperman and Ingenito [J. Acoust. Soc. Am. 67, 1988–1996 (1980)]. Their expressions for the correlation integrals were adapted to compute the correlation function for both simple and realistic propagation models, through the use of analytical and numerical Greens functions, respectively. The line source model involved the assumption of uncorrelated noise sources along an infinite line and was designed to model a shear zone in a shallow ice‐covered ocean. Using the corresponding Greens functions for a waveguide, the line model correlation function could be reduced to a summation over all mode pairs, each term of which involved an integration over wavenumber. In this paper, the line source correlation mode...

Signal field matching in range, depth, and bearing

Advances in propagation modeling have enabled increasingly accurate calculations of the acoustic field for situations where fairly complete knowledge of the acoustic environment is available. To take advantage of the improved propagation models, various matched field signal processing schemes have been developed. Because of the large number of possible combinations of range, depth, and bearing for a moving acoustic source, it is desirable to speed up the matching process and reduce the search region. This can be achieved by using an initial coarse global grid search followed by a fine optimization of source position to produce field matches. Further gains can be obtained by predicting the next source position so that subsequent searches can be restricted to small regions around the predicted positions. In this presentation, the essence of the normal mode propagation model and suitable processing and prediction schemes are described and applied to a shallow water environment. This is followed by illustrati...

Signal coherence in shallow water with rough boundaries

A normal mode model to predict signal coherence for a source moving in shallow water with rough boundaries has been described previously [J. M. Ozard, G. H. Brooke, M. J. Wilmut, and M. V. Greening, J. Acoust. Soc. Am. Suppl. I 69, S19 (1981)]. The model assumes that mode amplitude or mode phase fluctuate. We have extended this model to widely spaced sensors by permitting the fluctuations to be independent from sensor to sensor. This new model has been used to calculate signal coherence for a variety of sensor configurations. Coherences have been calculated for low frequencies for which only a few modes are present. When source receiver range is changing rapidly it is found that for closely spaced sensors coherence depends only on receiver separation, mode shape, and mode excitation. However, for widely spaced sensors or sources that maintain a constant source receiver range the roughness parameters have a profound effect on coherence. Sensor configurations that give consistently high signal coherence, an...

Coherent Noise Synthesizer

Abstract : A noise-generating algorithm and associated computer program for well-defined testing of beamformers are described. The algorithm is especially suitable for superdirective arrays of underwater hydrophones as it generates Gaussian noise of specified coherency. Statistical properties of the generator are confirmed to be those planned, and the ability of the generator to synthesize noise for isotropic or surface noise sources is verified for three-element arrays. Cumulative distributions for estimated coherency were obtained for the model. (Author)