D. Alexandrou

Angular dependence of 12-kHz seafloor acoustic backscatter

The angular dependence of seafloor acoustic backscatter, measured with a 12‐kHz multi narrow‐beam echo‐sounder at two sites in the central North Pacific with water depths of 1500 and 3100 m, respectively, has been determined for incidence angles between 0° and 20°. The acoustic data consist of quadrature samples of the beamformed echoes received on each of the 16 2.66° beams of a Sea Beam echo‐sounder. These data are subjected to adaptive noise cancelling for sidelobe interference rejection, and the centroid of each echo is determined. After corrections for the ship’s roll and raybending effects through the water column, the angles of arrival are converted to angles of incidence by taking athwartships apparent bottom slopes into account. For each beam, the mean echo power received is normalized by the corresponding insonified area that depends on the transmit and receive beam patterns, the ship’s roll angle and the local bottom slope. For lack of system calibration, the data are presented as relative mean...

Performance evaluation of multilayer perceptrons in signal detection and classification

Multilayer perceptrons trained with the backpropagation algorithm are tested in detection and classification tasks and are compared to optimal algorithms resulting from likelihood ratio tests. The focus is on the problem of one of M orthogonal signals in a Gaussian noise environment, since both the Bayesian detector and classifier are known for this problem and can provide a measure for the performance evaluation of the neural networks. Two basic situations are considered: detection and classification. For the detection part, it was observed that for the signal-known-exactly case (M=1), the performance of the neural detector converges to the performance of the ideal Bayesian decision processor, while for a higher degree of uncertainty (i.e. for a larger M), the performance of the multilayer perceptron is inferior to that of the optimal detector. For the classification case, the probability of error of the neural network is comparable to the minimum Bayesian error, which can be numerically calculated. Adding noise during the training stage of the network does not affect the performance of the neural detector; however, there is an indication that the presence of noise in the learning process of the neural classifier results in a degraded classification performance.

Adaptive noise canceling applied to Sea Beam sidelobe interference rejection

An adaptive noise cancelling (ANC) technique involving a joint-process deterministic least-squares lattice filter was applied to the Sea Beam bathymetric survey system data. The filtering scheme used in Sea Beam adversely affects the underlying acoustic return and may also lead to bathymetric artifacts. The authors investigate a possible remedy for this sidelobe interference problem offered by ANC coupled with signal preservation, provided both amplitude and phase information. The joint-process deterministic least-squares lattice is the adaptive filter of choice because of its superior transit response in the presence of power discontinuities. A REVGEN (reverberation generator) simulation (R.P. Goddard, 1985) of the Sea Beam system provided support for the proposed filtering technique. A complex data acquisition system was designed and built to record the in-phase and quadrature component of Sea Beam returns. Initial ANC processing of these recorded Sea Beam data provided satisfactory sidelobe interference cancellation with no noticeable degradation of the actual bottom returns. >

Full‐field optimum detection in an uncertain, anisotropic random wave scattering environment

The optimum detection of an unknown object in an uncertain random wave scattering environment is considered. A physics‐based approach to the design of the optimum detector is presented which merges statistical physical modeling of the acoustic scattering medium with a probabilistic description of environmental prior knowledge within a Bayesian decision‐theoretic framework. For the high‐frequency, shallow water, reverberation‐limited environment considered herein, the parametrization of the acoustic medium is essentially limited to modeling acoustic interaction with anisotropic seafloor microroughness with unknown horizontal wave‐number spectrum parameters. Simulation results, presented in terms of receiver operation characteristic (ROC) curves, aim to illustrate three principal points: (1) the cost of ignoring the bottom reverberation spatial coherence when it is present in the data; (2) the sensitivity of the likelihood ratio detector for a known environment to incorrect prior knowledge of the microrough...

Source localization in an uncertain acoustic scattering environment

The sensitivity of conventional matched‐field processing algorithms to uncertainty in ocean acoustic environmental parameters has prompted the design of more robust methods for source localization. In a recent study, Richardson and Nolte [J. Acoust. Soc. Am. 89, 2280–2284 (1991)] reported on the development of a new algorithm which incorporates prior knowledge of the environmental uncertainty into the design of the matched‐field processing algorithm. The result was the optimum uncertain field processor (OUFP). The present study addresses the problem of source localization in an imperfectly known acoustic scattering environment. The propagation scenario of interest, the surface duct, is characterized by the property that the received pressure field consists primarily of rays which are scattered from the sea surface. The surface roughness statistics are presumed to be axisymmetric, parametrized by rms height and correlation length. Two approaches to modeling the scattered field will be considered (1) the Ec...

Boundary reverberation rejection via constrained adaptive beamforming

The problem of selective reverberation cancellation, whereby both ‘‘signal’’ and ‘‘noise’’ are constituent components of the received reverberation process, is the focus of this article. The proposed solution involves the application of a constrained adaptive beamforming technique. The ‘‘prewindowed’’ deterministic least‐squares lattice filter is used as the central adaptive element. Constraints are in the form of simple spatial filtering prior to adaptation. The spatial correlation characteristics of volume and boundary reverberation are found to be directly applicable in a reverberation cancellation context. Experimental verification is offered by processing reverberation data from a shallow‐water deployment of an active sonar system. It is shown that the boundary reverberation components can be effectively suppressed while preserving the volume return. Computer simulations of the experiment offer additional insight into the adaptation process.

Application of high-resolution beamforming to multibeam swath bathymetry

In the context of swath bathymetry with multibeam echo-sounders, seafloor echoes received at regularly spaced elements of a hydrophone array are summed coherently to form a number of directional beams from which athwartships depth measurements are derived. This process can be implemented as a conventional beamformer leading to estimates of the direction of arrival of the echoes for each time sample. The process is inadequate in resolving closely spaced synchronous returns and the accuracy of these estimates is proportional to the number of acoustic data samples used in the process. To improve the angular resolution the authors have considered a number of high-resolution algorithms well known in power spectral estimation applications: autoregressive techniques (i.e. Yule-Walker, and unconstrained least squares), minimum variance methods (i.e. Capons method), and eigenanalysis algorithms (i.e. MUSIC). Comparisons of results obtained with realistic multibeam sonar simulations show that these algorithms have higher accuracy and better potential for high-resolution bathymetry than the conventional beamformer under nominal SNR levels.<<ETX>>

ROC evaluation of adaptive beamforming in a simulated shallow water environment

An optimal evaluation of adaptive beamforming techniques in a reverberation-limited shallow water environment is presented. A comprehensive simulation, using the sonar simulation toolset (SST) software in conjunction with the generic sonar model (GSRT) software, is used to create realistic beam data complete with target, noise, and reverberation. Adaptive beamforming techniques from the recursive least squares (RLS) family are applied to enhance detection performance via interference rejection. Two techniques are considered: linearly constrained beamforming using the minimum variance distortionless response (MVOR) beamformer and constrained adaptive noise cancelling (ANC) using the joint process least squares lattice (JPLSL) algorithm. Target detection trials, summarized in the form of receiver operator characteristics (ROC), are used to evaluate performance of the two adaptive beamformers. Results demonstrate mixed performance in reverberation-limited shallow water environments.

ROC performance evaluation of multilayer perceptrons in the detection of one of M orthogonal signals

A neural network detector is compared to an optimal algorithm from signal detection theory for the problem of one of M orthogonal signals in a Gaussian noise environment. A receiver operator characteristics (ROC) analysis is used. The neural detector is a multilayer perceptron trained with the backpropagation algorithm, while the optimal detector operates based on a likelihood ratio test. It was observed that for the signal-known-exactly case (M=1) the performance of the neural detector converges to the performance of the ideal Bayesian decision processor; however, for a higher degree of uncertainty (i.e., for a larger M) the performance of the multilayer perceptron is obviously inferior to that of the optimal detector. In addition, it was concluded that noise information in the training stage affects only slightly the performance of the neural detector. However, the knowledge of the noise distribution proved to be vital for the detection theory processor.<<ETX>>

Adaptive beamforming with the multichannel least squares lattice using beam-space constraints

Rejection of reverberant interference in active sonar systems via adaptive beamforming is approached by means of adaptive noise canceling (ANC). The multichannel version of the joint process least squares lattice algorithm (JPLSL) is the selected adaptive engine by which interfering reverberation is eliminated from beamformed data. Through application of beamspace constraints, signals arriving from selected directions are protected from cancellation by the JPLSL filter. Simulated data, generated by the Sonar System Toolset (SST) software, are used to demonstrate adaptive beamformer performance.<<ETX>>