Search

Describes a novel approach to the extraction of geometric features from sonic data. As is well known, a single measurement using a standard POLAROID sensor, though yielding relatively accurate information regarding the range of a reflective surface patch, provides scant information about the location in azimuth or elevation of that patch. This lack of sufficiently precise localization of the reflective patch hampers any attempt at data association, clustering of multiple measurements or subsequent classification and inference. The problem is particularly apparent in uncertain environments with unknown geometry, such as is found underwater. Moreover, the underwater environment precludes the usual (office-environment) simplification of two-dimensionality. The authors propose a multi-stage approach to clustering which aggregates sonic data accumulated from arbitrary transducer locations in an on-line fashion. It is computationally tractable and efficient despite the inherent exponential nature of clustering, and is robust in the face of noise in the measurements. It therefore lends itself to applications where the transducers are fixed relative to the mobile platform, where remaining stationary during a scan is both impractical and infeasible, and where deadreckoning errors can be substantial. The approach may be used both for map-building during exploration and for feature identification during navigation. >

This paper describes a non-coherent correlation technique for the formation of focused acoustic images, mainly for underwater applications. The technique is applied to the envelope of signals received by an array. The performances of the technique (lateral resolution, depth of field, accuracy of range determination, and speed of the imaging process) are estimated and compared with those of focused beamforming. The absence of grating lobes, for whatever arrangement of the transducers, and a high speed are the two major advantages of the proposed technique. The results of some simulations are reported and compared with those obtained by the beamforming process. >

Discusses the application of the phase mode techniques to the processing of signals in a circular sonar array and in particular in the application of high resolution algorithms such as MUSIC and ESPRIT. Simulation and experimental results are presented. >

The source, receiver and target dynamics in an ocean waveguide have significant effects on the performance of high resolution spatial and temporal processing. In a multipath environment, the motion causes Doppler shifts that are different for different paths. A procedure is developed for element level time series simulation for high frequency sonar returns in a monostatic scenario that includes the effects of source, receiver and target dynamics. The extension to the bistatic case is straight forward. The procedure consists of propagating the signal from the source to the target using ray tracing, projecting the incident field on to the scattered field, and propagating back to the receiver. The source, receiver and target dynamics is taken into consideration during propagation from the source to the target and from target to the receiver. The projection of the incident field onto the scattered field is accomplished using a T-matrix formulation. The modeling procedure is applied to a shallow water ocean waveguide with a rigid sphere as a target. The source, receiver and target motion is confined to a constant depth without loss of generality. The results demonstrate different Doppler shifts for different paths and the received signal exhibit the effect of motion. >

The new generation low-frequency echo-sounders, covering the full angular range, are also able to record acoustic images comparable to side-scan sonars. Possibilities of using these systems for sea-floor type identification are currently being investigated. Experimental echoes obtained with a 13 kHz multibeam echo-sounder have been collected and processed for a variety of sea-floor types and water depths. The results of two different analysis methods are presented. The backscattering strength of the sea-floor and its angular variations are known to be strongly dependent on the seabed type. They may be obtained as a function of incident angle from the measured levels averaged in narrow tilted beams, after correcting for the insonified area, the refraction by the sound-speed profile and the sea-floor local slopes. The results of this method make clear large differences between the various investigated seabed classes. The spectral features of backscattered time signals have previously been shown to be a useful tool for identification by side-scan sonars. The authors present the first results obtained by this method for deep sea-floor multibeam echoes: notable discrepancies between the spectral parameters corresponding to various seabeds are shown, enhancing the possibility of using such a technique for low-frequency echo-sounders. >

The application of synthetic aperture for sonars is hindered by the major difference between the celerity of acoustic waves for sonars and the celerity of electromagnetic waves for radars. This difference entails a greater sensitivity of signal processing to phase instabilities due to the inhomogeous medium or due to the unsteadiness of the sonar carrier. Nevertheless, the constant progress achieved in computing power, in the precision of inertial stations and in signal processing, should lead to sonars with synthetic aperture arrays in a near future. In that respect the first experiments conducted by GESMA are encouraging. The characteristics of the synthetic aperture are described, as well as the problems related to its application to the sonar field. The infrastructure of measurements, set up for the acquisition of data from a sonar with a synthetic aperture array, the motion of which is perfectly known, is described. The experimental results and a number of images obtained with real targets from the measurements are shown. >

Analysis of underwater acoustics signals involves the computation of statistics parameters for detection and estimation purposes. Spectrogram representations of the time and frequency domains can detect and track tonal concentrations from possible targets that can be inferred from the spectral signature presented on the distribution; usually this signature is characterized by energy lines whose locations depends on the target motion. Recently, new approaches for time-frequency representations have been proposed offering higher resolutions. On this paper, new approaches based on adaptive time extensions are presented and compared with other high resolution time-frequency techniques. A qualitative analysis is carried out on a signal coming from a passive sonar and attention is focused on improvements over the spectrogram for the weakest samples of the signal. >

An important problem in active sonar is the classification of objects from their returned scattered acoustic pressure field. Since in active sonar the insonifying waveform is usually broadband the received pressure is processed by matched filtering. After matched filtering an object's spectrum often contains sufficient information for classification. However, when objects are similar motion modifies their spectra enough so that the probability of false classification increases. In this paper the object's motion is first extracted and then sequential classification proceeds based on its measured fourth-order cumulant spectrum for each active sonar return and known fourth-order cumulant spectra of objects with similar velocity. The fourth-order cumulant spectrum is used because of its unique properties of noise suppression and feature extraction. >

The TMS320 series SLSI (Super Large Scale Integration) used mainly in digital signal processing (DSP) has been developed by Texas Instruments. TMS320c25 is its second generation DSP SLSI, it has characteristics of fast speed, low power consumption and has been widely used in DSP system. TMS320c30 is its third generation production. Its characteristics are all-floating point operation, faster speed, larger internal storage and ease of programming. We have developed a sonar signal processing system based on TMS320c25 and TMS320c30. It consists of a digital filter, raising sampling rate system, DICANNE interference cancelling system, adaptive linear-spectrum enhancing system, calculating distance and direction system and FFT and ZOOM FFT system. The results of calculation are sent to display and expert systems to find a moving target. >

The detection performance for fusing the output of multiple acoustic sensors is considered. Three fusion methods are analyzed and compared: fuse by summing-N-signals; the M-of-N processor, based on 0,1 decisions at each sensor; and threshold-and-sum N threshold signals. The last process uses two thresholds: T/sub 1/ which sets a lower limit on signal amplitudes for processing; and T/sub 2/, the detection threshold for the N summed signals which are greater than T/sub 1/. Eliminating low SNR data decreases data rates in the processor, and in practical situations, can improve the detection performance of a system when compared to the above processors. >

The automatic recognition of noise situations in an underwater environment is an interesting issue, especially in coastal waters. In fact, recognition of ships by their acoustic signature captured by passive sonars can be very useful for automatic statistic tasks. This is a complex problem involving a large amount of signals with unstable characteristics. To achieve our goal, an artificial intelligence based approach was adopted in order to extract results from such large amounts of data. Traditional machine learning and neural network techniques were used and compared in their performance to select a technology able to produce a good classifier. Based on the neural network technology, a real-time classifier, using low cost hardware was developed. >

DELPH-SONAR is a digital recording and processing system for two side scan sonar channels. Radiometric and geometric corrections produce on output a true map of sonar images. Results obtained from high level processings such as pipeline detection, freespan measurement, seafloor classification and screen digitization are stored in database and can be integrated onto the geographical map. >

BBN Systems and Technologies (BBN) has developed a family of advanced acoustic processors (AAP) to support the company's contract research and development in shallow water sonar. The AAP family is built with commercial-off-the-shelf (COTS) hardware and uses BBN's own distributed data flow software environment. This paper presents an overview of the AAP hardware and software architectures. Several illustrative examples of applications of the AAP to both scientific and operational problems are also presented. >

The PASISAR very high resolution seismic data are recorded through a single channel streamer attached to a deep towed side-scan sonar, using conventional seismic sources located at the sea surface. This device provides an outstanding improvement of both horizontal and vertical resolutions of seismic profiles in deep waters. However, the motions of the towed fish and the dissymmetry of the acquisition device dramatically alter the quality of data. Therefore a dedicated processing is mandatory in order to provide the interpreters with legible seismic sections. >

It has been suggested that perhaps a multi-beam sonar would be a useful tool for studying the reverberation contours of the biologically active plankton regions of the ocean known as the deep scattering layer, DSL. A brief experiment, utilizing the SASS multi-beam sonar on board the US Navy's oceanographic survey ship USNS Tanner during sea trials in December 1992, was performed to evaluate this possibility. The results of an August 1993 analysis of the sonar data collected, and sound velocity profiles of the area, clearly indicated that SASS was capable of mapping the DSL. Having established SASS as a useful DSL investigative tool, more ambitious experiments were included in the Test Plan of the ocean survey ship USNS Wyman for its February 1994 sea trial period. In addition to collecting larger amounts of sonar data in different locations, observations of contour variations during the DSL's vertical migrations at sunrise and sunset were also made. This paper describes the DSL measurement system, the data collection process, the results of the experiments, and presents examples of post processed data plots generated using Mathcad 4.0. The SASS System could, with some not-too-difficult software modifications to existing bathymetric/image processing and contour display techniques, be made to perform DSL/Water Column mapping in real time. >

The increasingly stringent demands of signal and information processing applied to problems arising in underwater communications and elsewhere point to the need for filters that are free from internally generated noise arising from rounding effects in the calculations. The use of finite number systems in digital signal processing provides a means of overcoming or reducing this problem. A new transform employing the Mersenne numbers (Boussakta and Holt, 1992) is used to implement filter designs without rounding errors. Convolutions based upon this transform are presented to show both the responses of a range of filters and also the effects of different transform parameters. Attention is concentrated upon filters defined by multiple poles sets. The time responses of these filters are shown. The ease with which calculations exceeding the modulus may be detected with the new transform is demonstrated. Error-free filter implementations will be a valuable contribution to signal processing for many applications in underwater communications and sonar. >

High-resolution side-scan sonars perform high-quality sea-bed imaging. In these images, texture recognition is a reliable characterization of seafloor nature. This task is efficiently carried out by previously trained neural classifier. Once the classifier is chosen; it is trained to separate various classes of textures according to examples extracted from homogeneous regions labeled by an operator. Most of the classifiers are not trained directly on gray-level images, instead, the learning set consists of features vectors computed images. Generalization capabilities of neural networks ensure a correct classification of test images provided that they are close the training set examples. Unfortunately, poor performances are observed should the examples slightly differ from the training set or the validity range of the features is to narrow. In this paper, the authors present a hybrid neural classifier consisting in a supervised network driven by an unsupervised one. The original image is partitioned in small two-dimensional elements from which feature vectors are extracted. The vectors are clustered by a self-organizing feature map and each element is labeled according to its activity. A supervised neural network, suited for Incremental learning, is trained on these elements. The self-organizing map searches for occurrences of new textures by continuously analyzing features extracted form sonar images. When such event arises, incremental supervised networks update its knowledge. Several improvements stem from this classification paradigm. The unsupervised network provides a quantitative hint of feature-set discrimination ability, class-selection method ensures a strong coherence between the feature set and the texture classes and, finally, incremental learning reduces the initial training phase foregoing exploitation and allows on-line incorporation of additional knowledge. >

Systematic intensity variation with range across a sidescan sonar record, or "shading", degrades visualisation, and is particularly distracting when repeated in each record forming a sonar mosaic. Shade variation is difficult to predict, not only because of uncertainties in the instrument function, but also because backscatter variation with incidence angle, and hence range, depends on seabed material. The paper describes a method of shade correction in which the correction is computed from the statistics of the record itself. This is first segmented into a number of classes of seabed material using a statistical classification program, and then column histograms for each different type of material are equalized with each other by a table look-up procedure. The paper discusses the relevance of good segmentation to geometric as well as shade correction for the sonar mosaicing task. >

The paper addresses the problem of detecting echos from linear frequency modulated (LFM) active sonar transmissions when the propagation channel exhibits time spreading distortion (TSD) and fast fading distortion (FFD). Both distortion mechanisms reduce the performance of "high resolution" LFM replica correlators which are designed to improve detection performance in reverberation. Robust detectors are required in this situation. The paper compares two robust detectors: the segmented replica correlator (SRC) and the replica correlator followed by incoherent integration (RCI). The author shows that for LFM, (1) TSD and FFD have an equivalent effect on the signal, (2) the RCI and SRC have equivalent reverberation only output statistics, and (3) the RCI and SRC have equivalent performance against either TSD or FFD. The author points out that while virtually all robust active sonar detectors use the SRC, the RCI is preferable when the amount of distortion is not known apriori. >

Recent multi-beam echo-sounders have two functions: bathymetry for iso-depths processing and sonar imagery which represents the variability of the acoustic interface. These sonar data are processed to get a cartographic representation commonly named mosaic. The construction of this image is based on the geometrical acquisition of the signal and on image processing techniques. >

MOVIES-B software developed by IFREMER (Institut Francais de la Recherche pour l'Exploitation de la Mer), enables the recognition of echogram features as acoustic detection structures according to spatial-energetic contiguity criteria. These structures (shoals, plankton layers) are described by energetic, morphological, spatial and temporal descriptors. Specifying acoustic detections as individual shoals will allow shoal-by-shoal echo-integration and may lead to better biomass estimation and shoal species description. The conceptual and operational framework of MOVIES-B is presented in the first part of this paper. The second part presents a preliminary work conducted in order to assess the feasibility of classification and possible species identification of shoals using MOVIES-B software. >

This work deals with the response of underwater bottom insonified with wideband ultrasonic signals. The aim of the study is to investigate the feasibility to recognize the nature of sediment layer using a wideband sonar system. The processing of signals must take into account the wide frequency band of the backscattered echoes and so includes spectral analysis and time-frequency analysis. The experiments have been performed in a natural environment, in Lake Geneva. Bottom echoes have been collected on 5 different types of bottom. For each backscattered echo signal, 3 vectors of features have been extracted: directly from the amplitude of the impulse response spectrum or from the parametric modeling or from the time-frequency representation. The results of the classification by using a neural network are then analysed. >

The echolocation capabilities of dolphins and small whales exceed those of current man-made sonars. Dolphins, beluga whales and false killer whales can perceive small targets presented over 110 m away, can classify target shapes independent of internal target reverberation, can discriminate wall thickness differences in targets of less than .2 mm, and can operate in high noise environments. Recent natural observations indicate that several species may also detect and choose targets buried in sediment. These tasks are accomplished through the use of range-gated clicks that tend to be broad band with peak frequencies exceeding 100 kHz. The short (50 microsec) pulses can have amplitudes exceeding 220 db and bandwidths exceeding 60 kHz. This paper provides a short review of animal echolocation capabilities, methodologies used to examine them, and potential uses of neural networks and other signal processing techniques to understand and perhaps duplicate those animal capabilities. >

Analysis of signal transformations in the auditory periphery model, which includes first order neuron models, revealed several remarkable features which would be useful for a system of short acoustic pulses identification. This paper describes the model of the auditory periphery and the basic properties of pulses transformations within it. A description is also given of the identification system containing the periphery model as a preprocessing module as well as a block for reduction of signal description redundancy and for decision making. The model test results are also presented. >

The new IFREMER's Research Vessel THALASSA II devoted to fishing and oceanographic activities will be equiped with a lot of acoustic systems: fishing sounders and sonars, trawl instrumentation, multibeam echosounders, doppler profiler, etc.. According to the kind of cruise that will be done, several systems will operate at the same time. This analysis shows that time sharing will be a necessary procedure for handling the data, especially for filtering with asynchronous systems. >

A method for determining a statistic characterisation of sea bottom acoustic data in both shallow and deep water is introduced. This method is based on the consideration that the marine ambient noise is non uniform. Real acoustical digital signal processing methods optimised for Gaussian ambient noise may degrade significantly in a nonGaussian ambient noise, hence, it is important to characterise the biological and environmental acoustical noise accurately, before including its structure in the formulation of the algorithm. In fact it is proved that the ocean ambient noise (NOA) is plausibly non Gaussian (NOA-NG), through the third order statistics (or fourth order) analysis. Tabulated and graphic results derived by experimental data show a general consistency. The authors propose an alternative architecture of estimation for the acoustical measurements techniques in the marine environment. This work is essential in the acoustic estimation of fish behaviour and the connection with the sea bottom nature. >

In support of a research programme into multi-static active sonar systems, DRA Portland have funded the development by Signal Computing Ltd of an experimental sonobuoy, a data recording system, and real-time signal processing and display equipment. A receiver array with narrow band linear signal processing was required which could be economically manufactured in small quantities for trials use. This was achieved by replacing the electronics of a production sonobuoy to provide a low-noise pre-amplifier and 16-bit delta-sigma ADC for each hydrophone, and a general purpose DSP in the lower unit to reduce the bandwidth of data for transmission. Deployment on sea trials has demonstrated the effectiveness and flexibility of the system. To facilitate monitoring of data from multiple sonobuoys, a real-time beamforming and display system was developed. This made use of general purpose DSP cards to provide a high degree of flexibility and high performance in an economical system. >