Adaptive Beamforming for Submarine-Satellite Communications with the (MBCA) Multielement Buoyant Cable Array Antenna
Blair
Carlson
MIT Lincoln Laboratory
244 Wood Street
Lexington, MA 02420
tel: (781) 981-2875
email: bcarlson@ll.mit.edu
Abstract
In order to provide the capability for submarines to communicate through a
satellite while remaining submerged and traveling at operational speeds a towed
buoyant cable array antenna is being developed.
The array is adaptive from the point of view that the direction of the
satellite need not be known, the position and orientation of the array need not
be know, and the shape of the flexible array need not be known.
A blind equalization procedure is used to estimate the signal space from
the downlink signal and create a spatial matched filter for receive.
While the frequency division satellite system is intended to allow only
one signal per frequency slot, the system can also operate in the presence of
jamming by separating multiple sources spatially.
Once the downlink receive antenna weights have been obtained, the more difficult task of obtaining uplink weights at a separated frequency must be performed. Since no data is available for blind equalization at the transmit frequency a frequency extrapolation method is used to extend the downlink receive weights to frequencies beyond where the equalization data was collected. This extrapolation is complicated by 2-pi ambiguities of the measured phases as well as amplification of measurement errors in the extrapolation process. An algorithm has been developed that performs well.
RFI
and Mainlobe Jamming Mitigation for Multichannel Imaging Radars
Patrick
Bidigare
Veridian Systems
3300 Plymouth Road
Ann Arbor, MI 48105
tel: (734) 994-1200 ext 2792
email: patrick.bidigare@veridian.com
Abstract
A
synthetic aperture radar (SAR) requires a single channel antenna to produce an
image of a scene containing only stationary objects and clutter. A number of papers have motivated the use of multiple channel
antennas in SAR for simultaneous scene imaging and ground moving target
indication. Here we consider the
utility of multiple channels in SAR for mitigation of radio frequency
interference (RFI) and jamming in the mainlobe.
A typical approach to interference
suppression when multiple antenna channels are available is spatial-only
adaptive beamforming. Here, a
linear combination of receive channels is used to produce a receive antenna
pattern null in the direction(s) of the interference.
When the real aperture width of the SAR antenna is small or stand-off
distances are large, the width of the antenna null produced can be an
unacceptably large portion of the SAR scene.
To make
this rigorous, we can cast SAR imaging as a problem of estimating the radar
cross-section of each range/Doppler cell in the presence of thermal noise and
localized RFI. We consider the width of the region of range/Doppler cells
whose Cramer-Rao variance bound exceeds a given threshold. We show that this width is proportional to the azimuth
resolution of the radar, which in turn is inversely proportional to the length
of the real antenna aperture.
We show
that by using both spatial (multiple channels) and temporal (multiple pulses)
adaptive processing that we can produce a null width which is proportional to
the doppler resolution of the radar, which in turn is inversely proportional to
the length of the synthetic antenna aperture.
This null width is thus independent of antenna size and standoff range,
and is much narrower than that obtained by spatial-only beamforming.
We
present a candidate architecture for STAP-based RFI/jamming mitigation in SAR
and show some results against data collected by Veridian ERIM International's
DCS radar.
Performance Analysis of the Derivative Based Updating Method
Michael
Zatman
MIT Lincoln laboratory
244 Wood Street
Lexington, MA 20420
tel: (781) 981 2543
email: zatman@ll.mit.edu
Abstract In this paper the asymptotic performance of the derivative based updating (DBU) algorithm is presented, setting an upper bound on the performance that could be achieved by finite sample implementations. The DBU method for adaptive beamforming and STAP in non-stationary interference environments has been proposed by a number of authors. However, until now the performance of this method has not been analytically described.
The DBU algorithm estimates both the weight vector and its derivative at time t=0 from the data available. Expressions are obtained for the method's effectiveness at estimating both these quantities as a function of the non-stationarity of the interference environment.
Adaptive beamforming with a rotating array and bistatic STAP examples are used to demonstrate the capabilities and limitations of DBU.
Further Evaluations of STAP Tests in Compound-Gaussian Radar Clutter
James H. Michels
AFRL/SNRT
26 Electronic Parkway
Rome, NY 13441
tel: (315) 330-4432
email: michelsj@rl.af.mil
Muralidhar Rangaswamy
ARCON Corporation
260 Bear Hill Road
Waltham, MA 02451-1080
tel: (781) 890-3330 x-226
email: murali@arcon.com
Braham Himed
AFRL/SNRT
26 Electronic Parkway
Rome, NY 13441
tel: (315) 330-2551
email: himedb@rl.af.mil
Abstract: The performance of a parametric space-time adaptive processing (STAP) method, the normalized parametric adaptive matched filter (N-PAMF), is presented here. Specifically, we consider signal detection in additive disturbance containing compound-Gaussian clutter plus additive Gaussian thermal white noise. Performance is compared to the normalized adaptive matched filter (NAMF) and the Kelly GLRT receiver using simulated and real data. We focus on the issues of detection and false alarm probabilities, CFAR, robustness with respect to clutter texture power variations and reduced training data support.
The N-PAMF test described in this paper requires no `a priori' knowledge of the disturbance statistics. This feature is important in real-time applications where such information is lacking. We also examine performance versus data support size used for disturbance estimation. This issue is of considerable importance for non-homogeneous clutter where representative secondary data cells are limited to a small set. Furthermore, the N-PAMF with a low model order facilitates significant detection performance improvement and training data support reduction over the Kelly GLRT and NAMF.
Impulsive Noise Mitigation in
Spatial and Temporal Domains for Surface-Wave Over-the-Horizon Radar
Yuri Abramovich and P. Turcaj
Cooperative Research Centre for Sensor Signal and Information Processing
SPRI Building
Technology Park Adelaide
SA 5095
Mawson Lakes, Australia
tel: +61 8 8302 3937
email: yuri@cssip.edu.au
Abstract The Surface-Wave Over-the-Horizon radar, due to its relatively long coherent integration time (up to several minutes) can be strongly affected by impulsive noise. In our case, the problem is compounded by the relative close proximity to the equator in Northern Australia. The impulses can be of three main origins - man made (external), natural (storms) or internal digital noise. Even a few sweeps affected by impulsive noise can significantly degrade signal-to-noise ratio and render the whole dwell useless.
In our paper, we will present a new and relatively simple method of detecting affected sweeps. Our digital receiver allows us to access backscattered signal free ranges. After range processing these ranges do not contain any signal of interest nor clutter generated by it. This can be used to our advantage as the only signal which stands out of the noise floor is impulsive noise. This makes it relatively easy to detect the sweeps of interest.
Our next step is to mitigate the impulsive noise in such a way that the overall effect is limited. The current state of the art is based on two main methods - temporal or spatial mitigation. The temporal method tries to replace whole sweeps by some optimal ones in order to limit the overall corruption of the dwell. The spatial method uses an antenna pattern side lobes reduction algorithm. Both of the mentioned methods loose effectiveness with the increasing number of sweeps contaminated by impulsive noise.
We propose method which combines the spatial and temporal ones based on the properties of the impulsive noise and improve the overall impulsive noise mitigation. The results are demonstrated on real data collected over several months in our experimental facility near Darwin, Australia.
Investigation of Methods for Mitigation of Tow Ship Noise for Volumetric Towed Arrays
Herbert Freese
1710 SAIC Drive
McLean, VA
tel: (703) 676-4743
email: herbert.a.freese@saic.com
John Ianniello
email: ianniellojp@npt.nuwc.navy.mil
Abstract not available
Presentation not available
Subband Energy Detection Methods in
Passive Array Processing
Michael Bono
Applied Research Laboratories
The University of Texas at Austin
10,000 Burnet Road
P.O. Box 8029
Austin, TX 78713-8029
tel: (512) 835-3033
email: mbono@arlut.utexas.edu
Roy Bethel
MITRE
Pete McCarty
ARL:UT
Ben Shapo
DSR
Abstract not available
Presentation not available
Case of a Short Data Record for Both Training and Signal Detection
Bldg 2187, CST 5
48110 Shaw Road
Patuxent River, MD 20670
tel: (301) 342-9104
email: freburgerbe@navair.navy.mil
D.W. Tufts
Kelley Hall
University of Rhode Island
Kingston, RI 02881
Abstract
The most difficult case for adaptive detection is a scenario in which the
interference is so non-stationary that the same data record must be used for
both interference training and signal detection. Methods that work well with
ample training data may not always hold their performance properties under these
limiting conditions.
The methods of Cross Spectral Metric (CSM), Multistage Wiener Filter (MWF) and Principal Components Inverse (PCI) are compared in various scenarios for cases of interference data in which one is testing for a signal that may or may not be present in a snapshot within a given set of data. We also consider the more standard case in which the interference data is assumed, correctly, to be signal free and separate from the data vector to be tested for signal presence. One of the key considerations in the performance of these methods is that PCI is a power or subspace based method; whereas, CSM is a correlation or eigenvector based method. The MWF can be considered as a method of compromise between power and correlation. Subspace estimates can remain stable even though the basis vectors that describe that space may not. These facts lead to degraded performance of the CSM method when training data is limited and when signal is within the data set. The MWF is also affected to a lesser degree. Limited training data also has significant impact on the performance of the methods as a function of rank. Although CSM and the MWF have benefits for underestimated rank, the methods suffer in performance when the rank is overestimated. The adverse properties of each method are often hidden with large amounts of training data.
Adaptive Detection in Compound-Gaussian Noise via Recursive Estimation of the Covariance Matrix
Ernesto
Conte and Antonio De Maio
Dipartimento di Ingegneria Elettronica e
Università
di Napoli Federico II
Via
Claudio 21, 80125 Napoli, Italy
tel:
+39-081-7683152
e-mail: conte@unina.it
e-mail: a.demaio@unina.it
Giuseppe
Ricci
Dipartimento
di Ingegneria dell’Innovazione
Università
di Lecce
Via Monteroni, 73100 Lecce, Italy
e-mail: giuseppe.ricci@unile.it
Abstract not available
Presentation not available
Optimized Target Detection and Identification Using Full-Polarization Radar Waveforms
David Garren, Michael K. Osborn, Anne C.
Odom, and J. Scott Goldstein
Science Applications International
Corporation
4001 Fairfax Drive, Suite 675
Arlington, Virginia 22203
tel: (703) 248-7759
email: david.a.garren@saic.com
S. Unnikrishna Pillai
Polytechnic University
Six Metrotech Center
Brooklyn, NY 11201
Joseph R. Guerci
Defense Advanced Research Projects
Agency
Special Projects Office
Arlington, VA 22203
Abstract Pillai,
Oh, Youla, and Guerci [1,2] have developed a theory for calculating the optimum
multiple-channel transmission pulse shape and receiver response corresponding to
the target response matrix and the spectral characteristics of the noise and
clutter. The new discrete-time implementation of this method for
optimizing the multiple-channel transmission waveform to maximize the
signal-to-interference plus noise ratio for optimal target detection is similar
to that involved in maximizing the Mahalanobis distance between two target
echoes so as to optimize target identification discrimination.
This analysis investigates optimal multiple-channel waveform transmission and reception for the case of a single full-polarization waveform, i.e., one containing both horizontal and vertical polarization components. Specific numeric examples are presented for the detection of the T-72 main battle tank and the identification discrimination between the T-72 and the M1 tank. The resulting optimized full-polarization transmission waveform typically focuses most of its energy into a narrow frequency band corresponding to the maximum full-polarization target response at that aspect angle. The performance improvement of transmitting the optimized VHF-band full-polarization waveform over that of transmitting a full-polarization chirped waveform having the same total energy and duration yields a SINR gain of 5-8 dB, depending upon the relative target aspect angle. Transmission of the full-polarization waveform optimized for identification discrimination distance between the T-72 and the M1 gives an improvement of 1-5 dB in the Mahalanobis over that obtained from the transmission of a full-polarization chirped waveform having the same energy and duration.
Constrained Maximum Likelihood
Covariance Estimation for Time-Varying Sensor Arrays
David Rieken and Daniel R. Fuhrmann
Washington University in St. Louis
Dept. of Electrical Engineering
Campus Box 1127
St. Louis, MO 63130
tel: (314) 935-7551
email: rieken@essrl.wustl.edu
Abstract We examine the problem of maximum-likelihood covariance estimation using a sensor array in which the relative positions of the individual sensors change over the observation interval. We propose incorporating information about the changing array geometry to reduce the dimension of the search space. Since what is desired is an estimate of the covariance matrix at each sample time, we consider the set of all Hermitian matrix sequences of the same size. It can be shown that this space is a vector space. We will also show that any valid covariance matrix sequence will lie within a vector subspace W1 and a convex set W2. Then the parameter to be estimated is in the intersection of W1 and W2 (which forms a cone), and it is necessary only to search that space for a maximum-likelihood solution.
The steps in determining the ML estimate are as follows: 1) Find an
orthonormal basis for W1. 2) Construct a rough estimate of the covariance matrix
sequence to be used as a starting point. 3) Project the sequence onto the
intersection of W1 and W2 using the iterative method of projection onto convex
sets (POCS). 4) Using the results of the previous step as a starting
point, perform a maxima search on the likelihood function using the orthonormal
basis for W1 and staying within W2. We will present an analysis of the ML
estimator as well as simulation results.
Resolution of Mainlobe and Sidelobe Detections
Using Model Order Determination
Amin G. Jaffer, Joe C. Chen, and Thomas W. Miller
Raytheon Electronic Systems
220 E. Imperial Highway
RE, Bldg. R7, M/S P527
P.O. Box 902
El Segundo, California 90245-0902
email: ajaffer@west.raytheon.com
Abstract This work presents the development and performance evaluation of a methodology for distinguishing between mainlobe and sidelobe detections that arise in adaptive radar systems operating in adverse environments. Various adaptive detection test statistics such as the adaptive matched filter (AMF), the generalized likelihood ratio test (GLRT) and the adaptive coherence estimate (ACE), and combinations of these, have been previously analyzed with respect to their sidelobe detection capabilities. In contrast to these methods that are based on detecting a single target with known direction and Doppler, the present method uses model order determination techniques applied to the AMF or GLRT data observed over the range of unknown angle and Doppler parameters. The determination of model order, i.e. the number of target signals present in the data, is made by successively fitting models composed of one signal, two signals, etc. to the AMF or GLRT data in a weighted least-squares sense. The corresponding residuals are used in the Akaike Information Criterion (AIC) to deduce the correct model order. Because we are concerned with resolving "false" sidelobe detections from true mainlobe ones (which tend to be separated by at least a beamwidth), the peaks of the AMF data provide reasonably accurate estimates of the angle parameters. These parameters are then held constant in the weighted least-squares model fits, thus avoiding a multidimensional angle search for the higher model orders and resulting in a computationally tractable algorithm.
Comprehensive Monte-Carlo computer simulation results are presented, including the probability of detecting one and two targets and the probability of false sidelobe detections for the AMF, GLRT methods and the method proposed here. These simulation results demonstrate substantial improvement in sidelobe rejection performance of the present method compared to previous methods.
Space-Time Adaptive FIR Filtering with Staggered
PRI
Richard Klemm
FGAN-FHR
Neuenahrer Str. 20
D 53343 Wachtberg, Germany
tel: +49 228 9435 377
email: r.klemm@fgan.de
Abstract
Space-time adaptive processing (STAP) has been recognized as an indispensable
tool for detection of moving targets by air- or space borne MIT radar. Future
military observation systems will include combined Synthetic Aperture Radar (SAR)
and ground moving target indication (GMTI) modes. GMTI techniques are in general
based on STAP techniques. One of the major issues of implementing STAP
techniques is the complexity of STAP architectures. A great deal of papers have
been published on subspace techniques which reduce the amount of operations
required for adaption, filter calculation and filtering. Some of these
techniques are based on rank reduction of the space-time covariance matrix,
others on the principle of order reduction. Space-time FIR filters belong to the
class of order reduced processors. They have been proven to achieve near optimum
clutter cancellation performance at dramatically reduced number of operations
[1, 2, 3]. Real-time processing capability has been proven by implementation on
a Mercury multi-processor system.
Besides a few disadvantages concerning the implementation the choice of a staggered PRI code offers some significant advantages: 1. Unambiguous estimation of the target Doppler; 2. Elimination of blind velocities; 3. Reduction of the threat posed by spot jammers. In this paper the effect of staggered PRI on STAP FIR filtering is discussed. It is shown that 1.the optimum fully adaptive processor based of staggered PRI works almost perfectly; 2. a FIR filter with constant coefficients suffers considerable losses compared with operation at constant PRI; 3. A reasonable clutter rejection performance is achieved be readapting the filter at each PRI.
As far as implementation is concerned readapting the STAP FIR filter at each PRI does not pose a significant problem since the number of filter coefficients can be chosen very small. Therefore, the number of training samples and hence the number of operations required for adaption is small too so that the filter coefficient can easily be calculated using the data received during one PRI.
In the paper a comparison of the performance of the STAP FIR filter using uniform or staggered PRI made. It comes out that staggering leads to losses in signal detection performance of a few dB.
Parametric Filters For Non-Stationary
Interference Mitigation In Airborne Radars
Peter Parker and A. Lee Swindlehurst
Brigham Young University
Dept. of Electrical & Computer
Engineering
459 CB
Provo, UT 84602
tel: (801) 378-4119
email: parkerp@ee.byu.edu
Abstract Multichannel
parametric filters are currently being studied as a means of reducing the
dimension of STAP algorithms for interference rejection in airborne
pulsed-Doppler radar systems. These filters are attractive to use due to
the low computational cost associated with their implementation as well as their
near optimal performance with a small amount of training data for a stationary
environment.
With the recent interest in circular antenna arrays as well as bistatic radar systems, several of the standard reduced dimension STAP algorithms have been modified to account for the changes in the clutter statistics across range bins. This paper will present a similar modification to the Space-Time AutoRegressive (STAR) filter that we previously proposed. This extension is based on the Extended Sample Matrix Inversion (ESMI) technique. Using a data set generated by MIT Lincoln Laboratory for a circular antenna array, we show that an improvement in performance over the standard STAR filter is achieved when the training data has range-varying statistics. This improvement is achieved with a similar computational cost and sample support for both of the filters. We also show that the modified STAR filter will yield a much greater SINR than the modified PRI-staggered STAP method for the same sample support.
An additional problem encountered in airborne radar systems is terrain scattered interference or hot clutter. A different modification to the STAR method is presented to account for a hot clutter environment. In addition to this, a three-dimensional
parametric approach is presented that will help mitigate mainbeam jamming signals. Using the same data set as above augmented with synthetic hot clutter we show that the extensions we present offer an increased SINR over the standard STAR filter as well as over the optimized pre-Doppler STAP algorithm.
Simulation and Analysis of Adaptive Interference Suppression for Bistatic Surveillance Radars
C. Frederick Pearson and Geordi
K. Borsari
MIT Lincoln Laboratory
244 Wood St
Lexington, MA 02173
tel: (781) 981-4118
email: pearson@ll.mit.edu
Abstract Adaptive algorithms providing interference suppression for airborne monostatic surveillance radars are by now reasonably well established. Recently, interest is growing in bistatic surveillance architectures. While these architectures potentially offer significant operational advantages, the bistatic geometry inherently leads to interference environments that vary along each of the range - doppler - azimuth axes ( in contrast, monostatic interference environments are usually stationary or quasi – stationary in the range dimension). The non-stationarity creates new challenges for adaptive estimation and suppression of the interference environment. In particular, traditional monostatic approaches to training the adaptive algorithms in the non-stationary bistatic environment will generally yield poor estimates and suppression of the interference in the region under surveillance. The technical goal is to develop new approaches to overcome these challenges in the bistatic geometry.
This talk reviews some of our recent analysis of bistatic surveillance systems, including both air- and space- based stand off transmitters. We first present some initial covariance analysis based studies, examining various proposed approaches to the bistatic interference suppression problem. We next present a full featured time series simulation model, incorporating high fidelity modeling of the terrain geometry and the front - end (non - adaptive) signal processing. We then consider various proposed bistatic algorithms – both partially model based and fully adaptive - in these realistic signal environments, and provide some preliminary assessments of relative performance.
Interference Estimation and Mitigation for STAP using the Two-Dimensional Wold Decomposition Parametric Model
Joseph M. Francos, Wenyin Fu,
and Arye Nehorai
University of Illinois at Chicago
EECS Department (M/C 154)
851 South Morgan Street 1120 SEO
Chicago, IL 60607-7053
tel: (312) 996-2778
email: nehorai@eecs.uic.edu
Abstract We develop parametric modeling and estimation methods for STAP data based on the results of the 2-D Wold like decomposition. The proposed parametric estimation algorithm of the interference components simplifies and improves existing STAP methods for target detection.
The 2-D Wold like decomposition implies that any 2-D regular and homogeneous discrete random field can be orthogonally decomposed into a sum of a purely-indeterministic field, a harmonic field and a countable number of mutually orthogonal evanescent fields. We show in this paper that the same parametric model that results from the above orthogonal decomposition naturally arises as the physical model in the problem of space-time processing of airborne radar data. In the space-time domain the target model is that of a harmonic component. The purely-indeterministic component of the space-time field is the sum of a white noise field due to the internally generated receiver amplifier noise, and a colored noise field due to the sky noise contribution. The presence of a jammer results in a barrage of noise localized in angle and distributed over all Doppler frequencies. In the space-time domain each jammer is modeled as an evanescent component whose 1-D modulating process is a white noise. The ground clutter results in an additional evanescent component of the observed 2-D space-time field. Its spectral support is along a diagonal line in the azimuth-Doppler plane.
We exploit the correspondence of the above mathematical and physical models to derive a computationally efficient algorithm for parametrically estimating the jamming and clutter fields. Having estimated the parametric models of the interference terms, their covariance matrix can be evaluated based on the estimated parameters. Moreover, the difficult problem of evaluating the rank of the low-rank covariance matrix of the interference is solved as a by-product of obtaining the parametric estimates. The weight vector is then obtained by projecting the desired response into the subspace orthogonal to the estimated interference subspace. The proposed estimation procedure provides estimates of the covariance functions of the interference terms even from the information in a single range gate, thus achieving high performance gain when the data in the different range gates cannot be assumed stationary.
Relationship Between Detection Algorithms for
Hyperspectral and Radar Applications
Nirmal Keshava, Stephen M. Kogon, and Dimitris
Manolakis
244 Wood Street
Lexington, MA 02474
tel: (781) 981-3344
email: keshava@ll.mit.edu
Abstract
With the advent of recent sensors, hyperspectral sensing has evolved into a
promising technique for ground reconnaissance.
Hyperspectral data consists of hundreds of contiguous radiometric
measurements collected passively from each pixel in the scene, and detection
capitalizes on the exploiting the difference between target and background
spectral signatures. Theses
differences may occur in both the shape and intensity of the spectrum. Many
detection methods in hyperspectral processing employ signal models commonly used
in radar even though it is an active sensor.
Starting from a common signal model, we discuss adaptive detection algorithms for hyperspectral data by outlining fundamental similarities and differences with radar. Starting with the physical mechanisms underlying active radar processing and passive remote sensing, we examine the relationship between hyperspectral data and coherent measurements for STAP radar. Crucial to this comparison is the parallelism between complex-valued data collected from a pulsed, multi-element radar and real, non-negative data from contiguous radiometric channels, and how these different paradigms still lead to comparable detector structures. While STAP radar coherently combines elements and pulses to produce a two-dimensional angle-Doppler output image at every range, hyperspectral data is comprised of radiance measurements that produce a collection of two-dimensional intensity maps of the same scene at each wavelength. Resolution in radar is driven by signal waveform parameters, as well as the aperture spacing and length, whereas resolution for hyperspectral sensing is a function of altitude and optics. Detection in both hyperspectral and STAP processing is inhibited by background clutter, but radar processing also addresses the presence of jammers.
We demonstrate our results through experiments with real data and discuss the fundamental applicability of adaptive radar signal models to detection in hyperspectral processing. Invariably, we demonstrate that despite their differences, adaptive hyperspectral and STAP processing essentially both strive toward optimal processing of multi-dimensional data.
Linear Feature Detection in SAR Images
Philippe
Courmontagne
ISEM
/ L2MP CNRS
Place
Georges Pompidou
83000
Toulon, France
tel: 33 494038950
email: philippe.courmontagne@isem.tvt.fr
Abstract Recently, a great deal in the literature has been dedicated to ship wake detection in Synthetic Aperture Radar (SAR) images. Indeed, it is well known that SAR are able to show ship wakes as lines darker or sometimes brighter than the surrounding sea. Most of the detection algorithms use the Radon transform. Indeed, when an image contains a straight line or a segment, its Radon transform exhibits a narrow peak if the line is brighter than it’s surrounding, and a trough in the opposite case. Thus, the problem in finding lines is related to detect these peaks and troughs in the transform domain. Another methods use the detection of both ships and ship wakes. Given that SAR images are affected by a granular, multiplicative noise (called speckle), most of these detection algorithms pre-filter the data in order to improve the visibility of the ship wakes.
We know that the application of the Radon
transform needs the computation of interpolated image in a rotated reference
system. In this paper, we propose a new method based on the wedding between the
Radon transform and a speckle filtering technique. We use this filtering
technique to compute the interpolations of the SAR image, in order to estimate
properly the signal of interest (the ship wake) in the rotated reference system.
This processing is called the stochastic matched filtering technique. It is
based on the signal expansion into series of functions with uncorrelated random
variables for decomposition coefficients. This corresponds to the Karhunen-Loeve
expansion in the case of a white noise. The chosen basis functions improve the
signal to noise ratio after processing. With this process, there is no more
sinusoidal curves corresponding to the speckle in the Radon domain, so the
detection of the peak (or trough) corresponding to the ship wake is improved.
Experimental results on ERS SAR images are presented and compared to those
obtained using some classical approaches.
Space-Time Adaptive Matched-Field Processing (STAMP)
Yung P. Lee
Science
Applications International Corporation
1710 SAIC Drive
McLean, VA 22102
tel: (703) 676-6512
email: yung@osg.saic.com
Abstract Space-time adaptive processing (STAP) is two-dimensional adaptive filtering employed for the purpose of clutter cancellation to enable the detection of moving targets. It has been a major focus of research activity in radar applications for which the platform is in motion, e.g., airborne or space-based systems. In this setting, an antenna sensor array provides spatial discrimination, while a series of time returns or pulses form a synthetic array that provides Doppler (velocity) discrimination.
The application of STAP for the mobile towed-array sonar system is non-trivial because of the complex multi-paths in the underwater environment. On the other hand, Matched-field processing (MFP) that uses a propagation code to predict the complex multi-path structure and coherently combines it to provide range/depth discrimination has been studied and demonstrated. MFP with a synthetic array (a series of snapshots) to estimate the source velocity and localize source in range and depth has also been demonstrated.
STAMP combines the adjacent-filter beamspace post-Doppler STAP and MFP to provide improved performance for the mobile multi-line-towed-array sonar applications. The processing scheme includes: transforming phone time snapshots into frequency domain, at each frequency bin forming horizontal beams to the directions of interest for each towed line, then combining signal in multi-towed-lines and adjacent Doppler bins and beams that cover the multi-path Doppler spread due to motion with adaptive MFP. A study of STAMP performance in the towed-array forward-looking problem will be discussed. In this problem, the own-ship signal and its bottom scattered energy can be treated as stationary interference with a moving target at constant speed within processing interval of a few minutes.
Passive Differential Matched-field Depth
Estimation of Moving Acoustic Sources
Department of Electrical and
Computer Engineering
Duke University
Box 90291
Durham, NC 27708
tel: (919) 660-5274
email: jk@ee.duke.edu
Abstract Conventional passive matched-field processing (MFP) techniques for estimating source depth and range are degraded by target motion and environmental mismatch. In large part, this stems from the fact that multipath replicas are formed as a sum of normal modes whose relative phases depend on horizontal wavenumber differences multiplied by source range. In reality, however, these modal phases are both time-varying because of source motion and uncertain because of environmental model mismatch. In this paper, passive differential matched-field depth estimation (DMFDE) is presented which exploits source motion to achieve greater robustness to environmental mismatch. This work extends previous space-time matched-field depth estimation methods derived by the authors for active sonar and radar problems, to the passive sonar case where the temporal waveform of the target is completely random. DMFDE uses a non-linear state-space formulation wherein the embedded propagation model is used only to predict changes in the modal phases from snapshot-to-snapshot under a set of hypothesized target motions. Sequential importance sampling (SIS) is then used to solve for a recursive estimate of the a posteriori probability surface with respect to target depth and range-rate. Because DMFDE models only modal phase changes due to target motion, which depend on horizontal wavenumber differences multiplied by small changes in source range between snapshots, it is much less sensitive to environmental mismatch in the propagation model. Preliminary results indicate that DMFDE outperforms conventional MFP in its ability to produce easily identifiable, unambiguous peaks for depth estimation with a vertical line array in a shallow-water channel. Further simulation results characterizing depth and range-rate estimation accuracy in a realistic Mediterranean ocean channel will also be provided in the paper.
Covariance Matrix Filtering for Adaptive Beamforming with Moving Interference
Johns Hopkins University
Applied Physics Laboratory
11100 Johns Hopkins Rd.
Laurel, MD 20723-6099
tel: (240) 228-4287
email: bruce.newhall@jhuapl.edu
Abstract An approach is developed for adaptive beamforming for mobile sonars operating in an environment with moving interference from surface shipping. It is assumed that the sound source of each ship is drawn from an ensemble of Gaussian random noise, but each ship moves at constant speed along a deterministic course. An analytic expression for the ensemble mean covariance is obtained. In practice the location, course, speed, mean noise level, and transmission loss of each interferer are not known with sufficient precision to use the modeled ensemble mean as a basis for adaptive beamforming. The problem is thus to accurately estimate the ensemble mean based on data samples. The analytic ensemble mean is not stationary, and thus is not well estimated by the sample mean. The ensemble of covariance samples consists of rapidly varying random terms associated with the emitted noise and more slowly oscillating deterministic terms associated with the source and receiver motion. The non-stationary ensemble covariance mean can be estimated by filtering out the rapidly varying noise while retaining the slow oscillatory terms. Performance of the filters can be visualized and assessed in the "epoch frequency domain ", the Fourier transform of the covariance samples. In this domain, higher bearing rates show up at higher frequencies. The traditional sample mean stimator retains only the zero frequency bin corresponding to stationary interference. Techniques that can identify and include the appropriate non-zero frequency contributions are better non-stationary estimators than the sample mean. Several such techniques are offered and compared. Simulations are invaluable in evaluating the filter performance, since the ensemble mean can be precisely calculated analytically in the simulation, and compared directly with the sample estimates. Simulations of adaptive beamformers using covariance filtering will be shown to yield improved robustness to shipping motion.
Broadband Adaptive Beamforming: Motion Mitigation in the Littoral Environment by Frequency Averaging
1710 SAIC Drive
McLean, VA 22102
tel: (703) 676-5975
email: greener@osg.saic.com
Abstract:
The work exploits the propagation characteristics of the shallow water
environment to devise an alternative formulation of Adaptive Beamforming (ABF)
for broadband signals. Conventional
narrowband ABF methods, based on averaging multiple snapshots over time for a
fixed frequency, have a limited ability to null multiple moving ships,
particularly for large aperture arrays. The
broadband method requires that the signals and interferers are broadband and
spread in vertical arrival angle; both are generally the case in a shallow water
environment. This method is less
sensitive to target-interferer motion than narrowband methods because it
requires less integration time. Further, the estimation process for the
covariance matrix averages hundreds of samples, as compared to the tens of
samples available using narrow band methods, producing better statistical
convergence.
The significant contribution is the expansion of the replica vectors in terms of a set of basis functions, which are independent of frequency. Adaptation is then applied to the basis functions. For individual frequencies, replica vectors are adapted by projecting the replica vectors into the subspace spanned by the adapted basis functions. The basis functions are plane waves with fixed horizontal wave number. Each plane wave is a valid replica over a range of frequencies, although for each frequency the replica corresponds to signals having a different vertical arrival angle. When the shallow water environment supports propagation over a wide range of vertical angles, the covariance matrix for the basis function may be developed for a correspondingly wide range of frequencies. The same technique can be applied to range-focused basis functions as well. A variant of the method for arrays with vertical aperture allows adaptation of the basis functions in azimuth; conventional beamforming in the vertical aperture supports the distinction of deep from shallow sources of noise.
Beamspace Adaptive Beamforming for Hydrodynamic Towed Array Self-Noise Cancellation
MIT Lincoln Laboratory
244 Wood Street
Lexington, MA 02420
tel: (781) 981-5341
email: vpremus@ll.mit.edu
Abstract The deleterious effects of flow-induced vibrations, or cable strum interference, on passive acoustic detection using towed hydrophone arrays have been well known since the mid-1960s. Recently, it has been shown that significant cable strum self-noise rejection is attainable using a time-domain, beamspace approach which employs an LMS weight update and a reference channel steered to non-acoustic k-w space. In this work, a case study of array self-noise performance using an adaptive beamformer (ABF) in conjunction with a towed array passive sonar processing system is presented.
Following a brief overview of the physics
underlying flow-induced towed array self-noise, ABF design factors that impose
undesirable limitations on self-noise cancellation performance are identified.
The ABF algorithm under consideration consists of a minimum variance
distortionless response (MVDR) beamformer that incorporates a white noise gain
constraint (WNGC) based on the scaled projection technique. It will be shown
that the principal limitation of this ABF with regard to cable strum rejection
is the fact that the scaled projection algorithm, as implemented, overly
constrains MVDR adaptation in the frequency band usually associated with cable
strum. A frequency-dependent, scaled projection WNGC is proposed which enables a
more freely adaptive MVDR beamformer within the strum band, while maintaining a
stiffer WNGC at higher frequency where off-MRA signal model mismatch may induce
unwanted signal suppression.
The optimization of the scaled projection
WNGC represents a balance between enhanced adaptivity for self-noise suppression
and robustness to mismatch-induced signal suppression. This balance is examined,
using both simulation and data analysis, for dominant sources of mismatch. In
the strum band, these sources are believed to be unmodeled acoustic multipath
propagation and broadband dispersion. Further, the tradeoff in optimizing the
WNGC simultaneously for passive acoustic narrowband and broadband processing and
display is considered. Performance results for several cases of towed array data
will be presented.
Adaptive
Range-Focused Beamforming in Passive Sonar
for the Detection of Near-Field
Sources
Stephen Kogon
MIT
Lincoln Laboratory
244
Wood Street
Lexington,
MA 02420-9108
tel:
(781) 981-3275
email: kogon@ll.mit.edu
Abstract
Large aperture arrays are often desirable because they can provide a large
amount of gain and fine resolution. In addition, long arrays allow for effective
range discrimination by exploiting wavefront curvature that varies with range.
Effective ranging is of great utility to a submarine in short-range "close
encounters." The primary trouble with range-focused beamforming is the fact
that far-field sources defocus when focusing a beamformer in the near-field. The
far-field source blurs over a substantial angular sector. Often, this far-field
source is very strong, as in the case of merchant ships, so that the defocusing
causes the spreading of significant interference energy over a number of beams.
This reduced performance can limit the detection of close-range contacts.
Adaptive
beamforming, however, can null far-field sources when focusing in the
near-field, exploiting the differences in array response between near and
far-field sources. Thus, the detection of weak, near-field sources can be
accomplished in the presence of strong far-field interference. In fact, through
effective nulling in range, an adaptive beamformer can potentially detect the
presence of a near-field source at the same bearing as a far-field interferer.
In this paper, we will provide a detailed data analysis of several adaptive
range-focused beamformers applied to experimental data collected with a large
aperture towed array. We will begin with a characterization of interference from
far-field sources which motivates the beamspace selection in order to
effectively capture the interference for nulling purposes. Another consideration
for the resulting beamspace algorithm is robustness to self-nulling. To this
end, several methods will be examined, including minimum-variance soft
constraints and white noise gain control methods, and demonstrated on
experimental data. The analysis will seek to quantify performance improvement in
terms of detecting near-field sources and discuss trade-offs that become
necessary for fielding such an algorithm in actual real-time systems.
Bistatic STAP Data Collection with a Space-Based Illuminator
AF Research Laboratory
Sensors Directorate (SN)
26 Electronics Parkway
Rome,
NY 13441
tel: (315) 330-2211
email: mark.davis@rl.mil
Abstract The ultimate sanctuary for an all weather surveillance system is a space-based radar (SBR), providing the warfighter an ability to see targets without terrain shadowing and without the necessity of crossing international borders. However, there are severe limitations to the affordability of space-based radars for small targets, high area coverage rates, and accurate target tracking. A bistatic adjunct to the SBR can overcome these difficulties, if a suitable concept of operations (CONOPS) can be determined. Two enabling technologies that make this approach feasible are the multiple beam receiver architecture, and bistatic space-time adaptive processing (STAP) target detection.
A bistatic surveillance radar must synchronize
the receiver to match the transmit beam, in order to maximize the system
efficiency. For a space-based illuminator, the area illuminated by the
transmitter is typically much larger than the receiver beam angular area. For
optimum area coverage rate, the receiver UAV should employ a multiple beam
receiver to match the area illuminated from space. Results showing an emphasis
on data efficient STAP algorithms will be presented.
In the SBR scenario, the range and Doppler
clutter spectrum is changing rapidly, reducing the number of training cells
available for STAP weight computation. An analysis of bistatic geometries and
survey of algorithms that conserve the number of data samples required for
adaptation will be presented.
Finally, this paper will summarize the design of an experiment to validate these system concepts. The experiment employs a commercial satellite radar as bistatic illuminator and a multichannel airborne receiver as a bistatic STAP data collection platform. Analysis of the CONOPS will be presented to validate the data collection objectives. And simulation of the bistatic clutter scenarios and challenges to current bistatic STAP algorithms will illustrate the need for a long-term development program.
UHF Multichannel Airborne Radar Data Collection
Northrop Grumman ESSS
P.O. Box 746 / MS 899
Baltimore, MD 21203
tel: (410) 765-2162
email: robert_t_craig@mail.northgrum.com
Abstract In
September 2000, Northrop Grumman flew an experimental UHF radar system on our
BAC 1-11 aircraft in order to collect IQ data suitable for AMTI and FOPEN GMTI
algorithm development and demonstration. The system included a small
transmit array, a separate 12-element linear receive array, 12 receive channels
including IF digital receivers, engineering displays, and RAID. The FOPEN
flights included controlled ground moving targets under foliage cover, Moving
Target Simulator, and repeaters. The AMTI flight included a Sabreliner
target and a barrage noise jammer that was cycling on and off. The flight
paths were designed so that the target Doppler traversed nearly the entire
unambiguous Doppler space. The target thus competed against thermal noise,
sidelobe clutter, and mainlobe clutter, in all cases both with and without
jamming.
This paper will describe the radar system and the tests, and will include early AMTI data analysis results.
Unphysical Violation Of Reciprocity Detection And Mitigation
Daniel W. Bliss
MIT Lincoln Laboratory
244 Wood Street
Lexington, MA 02420
tel: (781) 981-3300
email: bliss@ll.mit.edu
Abstract/Presentation not available for publication
Detection and Mitigation of Multiple Low Cost Low Complexity (LCLC) Electronic Counter Measure (ECM) Devices
Veridian ERIM International
P. O. Box 134008
Ann Arbor, MI 48113-4008
tel: (734) 994 1200 ext. 2845
email: sharma@erim-int.com
Abstract/Presentation not available for publication