Search

ABSTRACT. The authors investigated systematic error associated with endpoints of memory-guided actions performed in near and far reaching spaces. To accomplish that objective, the authors instructed 12 participants to initiate [...]

We consider the problem of calculating the posterior Cramer-Rao lower bound (PCRLB) for tracking in cluttered domains in which there can be both missed detections and false alarms. We present a novel formulation of the PCRLB in which we initially determine a bound conditional on the sequence of measurements available. We then create an unconditional bound as a weighted average of these conditional PCRLBs. This new bound is proven to be less optimistic than the standard formulation of the PCRLB for cluttered environments recently developed [1, 2] and will therefore better predict optimal estimator performance. At each stage, the conditional PCRLB must evaluate the effect of the uncertain measurements, and we extend previous work [2] to show that the measurement origin uncertainty manifests itself as a single information reduction factor (IRF) that is dependent on the number of measurements available. We also present some useful approximations when the false alarm rate is low. Simulations then consider the problems of 1) determining the CRLB for the point of impact of a ballistic missile, and 2) determining the PCRLB for tracking a nearly constant-velocity (NCV) target in a high clutter environment. In each case, we compare the new bound with the standard approach, and as expected the new CRLB/PCRLB can be seen to be less optimistic. Moreover, in case 1) we compare the new CRLB with a heuristic bound specially constructed for this problem, and a maximum likelihood estimator (MLE). The new bound both compares favorably with the heuristic bound, and shows close agreement with the performance of the MLE. The new bound is therefore an accurate predictor of filter performance in this case. In example 2) we demonstrate some interesting features of the new theory. Of particular interest we determine both precisely when the new bound will be significantly greater than the standard bound and when the two bounds will be virtually identical. This is useful in determining when the new approach, with its greater computational burden, should be preferred to the established approach. We conclude that the novel PCRLB formulation introduced herein represents an exciting development in the determination of RMSE performance bounds in cluttered environments.

Acoustic nodes, each containing an array of microphones, can track targets in x-y space from their received acoustic signals, if the node positions and orientations are known exactly. However, it is not always possible to deploy the nodes precisely, so a calibration phase is needed to estimate the position and the orientation of each node before doing any tracking or localization. An acoustic node can be calibrated from sources of opportunity such as beacons or a moving source. We derive and compare several calibration methods for the case where the node can hear a moving source whose position can be reported back to the node. Since calibration from a moving source is, in effect, the dual of a tracking problem; methods derived for acoustic target trackers are used to obtain robust and high resolution acoustic calibration processes. For example, two direction-of-arrival-based calibration methods can be formulated based on combining angle estimates, geometry, and the motion dynamics of the moving source. In addition, a maximum likelihood (ML) solution is presented using a narrowband acoustic observation model, along with a Newton-based search algorithm that speeds up the calculation the likelihood surface. The Cramer-Rao lower bound (CRLB) on the node position estimates is also derived to show that the effect of position errors for the moving source on the estimated node position is much less severe than the variance in angle estimates from the microphone array. The performance of the calibration algorithms is demonstrated on synthetic and field data.

The problem of optimal state estimation of linear discrete-time systems with measured outputs that are corrupted by additive white noise is addressed. Such estimation is often encountered in problems of target tracking where the target dynamics is driven by finite energy signals, whereas the measurement noise is approximated by white noise. The relevant cost function for such tracking problems is the expected value of the standard [H.sub.[infinity]] performance index, with respect to the measurement noise statistics. The estimator, serving as a tracking filter, tries to minimize the mean-square estimation error, and the exogenous disturbance, which may represent the target maneuvers, tries to maximize this error while being penalized for its energy. The solution, which is obtained by completing the cost function to squares, is shown to satisfy also the matrix version of the maximum principle. The solution is derived in terms of two coupled Riccati difference equations from which the filter gains are derived. In the case where an infinite penalty is imposed on the energy of the exogenous disturbance, the celebrated discrete-time Kalman filter is recovered. A local iterations scheme which is based on linear matrix inequalities is proposed to solve these equations. An illustrative example is given where the velocity of a maneuvering target has to be estimated utilizing noisy measurements of the target position.

Many target tracking subsystems have the ability to schedule their own data rates; essentially they can 'order' new information whenever they need it, and the cost is in terms of the sensor resource. But among the unmanaged schemes, uniform sampling, in which a new measurement is requested periodically and regularly, is the most commonly-used sampling scheme; deliberately nonuniform schemes are seldom given serious consideration. In this paper, however, we show that such schemes may have been discarded prematurely: a nonuniform sampling can have its benefits. Specifically, the nonuniform and uniform sampling schemes are compared for two kind of trackers: the probabilistic data association filter (PDAF), which updates its track based on a single scan of information at a time; and N-D assignment (an optimization-based implementation of the multi-hypothesis tracker (MHT)), in which the sliding window involves many scans of observations. We find that given the ground rule of maintenance of the same overall scan rate (i.e., the same sensor effort) uniform sampling is always optimal for the single-scan tracker in the sense of track life. However, nonuniform sampling can outperform uniform sampling if a more sophisticated multi-scan tracker is used, particularly when 1) the target has a high process noise, and/or 2) the false alarm density is high, and/or 3) the probability of detection is high.

This study is concerned with multi-target tracking (MTT). The Cramer-Rao lower bound (CRB) is the basic tool for investigating estimation performance. Though basically defined for estimation of deterministic parameters, it has been extended to stochastic ones in a Bayesian setting. In the target tracking area, we have thus to deal with the estimation of the whole trajectory, itself described by a Markovian model. This leads up to the recursive formulation of the posterior CRB (PCRB). The aim of the work presented here is to extend this calculation of the PCRB to MTT under various assumptions.

We present a (suboptimal) filtering algorithm for tracking a highly maneuvering target in a cluttered environment using multiple sensors. The filtering algorithm is developed by applying the basic interacting multiple model (IMM) approach and the probabilistic data association (PDA) technique to a two sensor (radar and infrared, for instance) problem for state estimation for the target. A simultaneous measurement update approach is followed where the raw sensor measurements are passed to a central processor and fed directly to the target tracker. A multisensor PDA filter is developed for parallel sensor processing for target tracking under clutter. A past approach using parallel sensor processing has ignored certain data association probabilities leading to an inaccurate implementation. Another existing approach applies only to nonmaneuvering targets. The algorithm is illustrated via a highly maneuvering target tracking simulation example where two sensors, a radar and an infrared sensor, are used. Compared with an existing IMM/PDA filtering algorithm with sequential sensor processing, the proposed algorithm achieves significant improvement in the accuracy of track estimation.

This paper develops centralized and distributed tracker performance models that account for the fading detection performance that is common in active sonar contact data. We apply centralized and distributed tracking algorithms to simulated active sonar data that exhibits detection fading, and compare tracking performance to that predicted by the tracker models. We find qualitative agreement of model-based and simulation-based results, and we identify a tradeoff between centralized and distributed approaches: the former achieves better localization accuracy, while the latter achieves better performance in a receiver operating characteristic (ROC) curve sense.

Initializing and maintaining a track for a low observable (LO) (low SNR, low target detection probability and high false alarm rate) target can be very challenging because of the low information content of measurements. In addition, in some scenarios, target-originated measurements might not be present in many consecutive scans because of mispointing, target maneuvers, or erroneous preprocessing. That is, one might have a set of noninformative scans that could result in poor track initialization and maintenance. In this paper an algorithm based on the expectation-maximization (EM) algorithm combined with maximum likelihood (ML) estimation is presented for tracking slowly maneuvering targets in heavy clutter and possibly noninformative scans. The adaptive sliding-window EM-ML approach, which operates in batch mode, tries to reject or weight down noninformative scans using the Q-function in the M-step of the EM algorithm. It is shown that target features in the form of, for example, amplitude information (AI), can also be used to improve the estimates. In addition, performance bounds based on the supplemented EM (SEM) technique are also presented. The effectiveness of new algorithm is first demonstrated on a 78-frame long wave infrared (LWIR) data sequence consisting of an F1 Mirage fighter jet in heavy clutter. Previously, this scenario has been used as a benchmark for evaluating the performance of other track initialization algorithms. The new EM-ML estimator confirms the track by frame 20 while the ML-PDA (maximum likelihood estimator combined with probabilistic data association) algorithm, the IMM-MHT (interacting multiple model estimator combined with multiple hypothesis tracking) and the IMM-PDA estimator previously required 28, 38, and 39 frames, respectively. The benefits of the new algorithm in terms of accuracy, early detection, and computational load are illustrated using simulated scenarios as well.

We present in this correspondence an improved sequential Monte Carlo (SMC) filter for ballistic target tracking with random, time-varying ballistic coefficient. The proposed tracker is a sampling/importance resampling (SIR) filter that uses an optimized importance function to combat particle degeneracy, and also incorporates an additional measurement-driven Markov chain Monte Carlo (MCMC) move step to prevent particle impoverishment. Simulation results show that, using significantly fewer particles than previously reported in the literature for similar tracking problems, the root mean-square error (RMSE) curves for the proposed optimized SIR filter approach the square root of the ideal posterior Cramer-Rao lower bound (PCRLB).

Two equations express detection probability and detectability factor for all swerling targets and for partially correlated intermediate cases.

A hidden Markov model (HMM)-based method for recognizing aerial targets according to the sequential high-range-resolution (HRR) radar signature is presented. Its recognition features are the location information of scattering centers extracted from the HRR radar echoes by the Relax algorithm. The HMM is used to characterize the spatio-temporal information of a target. Several HMMs are cascaded in a chain to model the variation in the target orientation and used as classifiers. Computer simulations with the inverse synthetic aperture radar (ISAR) data are given to demonstrate that for an open-set recognition, average class-recognition rates of 84.50% and 89.88% are achieved, respectively, under two given conditions.

Hybrid trackers incorporating a non-Markov modal model have been proposed. However, these algorithms use a restricted measurement set to identify the modal state. It is shown here that a renewal process model, along with broader measurement fusion, enhances modal classification in such estimators.

A novel solution is provided for the bias estimation problem in multiple asynchronous sensors using common targets of opportunity. The decoupling between the target state estimation and the sensor bias estimation is achieved without ignoring or approximating the crosscovariance between the state estimate and the bias estimate. The target data reported by the sensors are usually not time-coincident or synchronous due to the different data rates. Since the bias estimation requires time-coincident target data from different sensors, a novel scheme is used to transform the measurements from the different times of the sensors into pseudomeasurements of the sensor biases with additive noises that are zero-mean, white, and with easily calculated covariances. These results allow bias estimation as well as the evaluation of the Cramer-Rao lower bound (CRLB) on the covariance of the bias estimate, i.e., the quantification of the available information about the biases in any scenario. Monte Carlo simulation results show that the new method is statistically efficient, i.e., it meets the CRLB. The use of this technique for scale and sensor location biases in addition to the usual additive biases is also presented.

For a multi-sensor target tracking system, the effects of temporally staggered sensors on system performance are investigated and compared with those of synchronous sensors. To capture system performance over time, a new metric, the average estimation error variance (AEV), is proposed. For a system that has N sensors with equal measurement noise variance, numerical results show that the optimal staggering pattern is to use N uniformly staggered sensors. We have also shown analytically that the AEV of the system with N uniformly staggered sensors is always smaller than that of the system with N synchronous sensors. For sensors with different measurement noise variances, the optimal staggering pattern can be found numerically. Practical guidelines on selecting the optimal staggering pattern have been presented for different target tracking scenarios. Due to its simplicity, uniform staggering can be used as an alternative scheme with relatively small performance degradation.

A new class of variable-structure (VS) algorithms for multiple-model (MM) estimation is presented, referred to as expected-mode augmentation (EMA). In the EMA approach, the original set of models is augmented by a variable set of models intended to match the expected value of the unknown true mode. These models are generated adaptively in real time as (globally or locally) probabilistically weighted sums of mode estimates over the model set. This makes it possible to cover a large continuous mode space by a relatively small number of models at a given accuracy level. The paper presents new theoretical results for model-set design, a general formulation of the EMA approach, along with theoretical analysis and justification, and three algorithms for its practical implementation. The performance of the proposed EMA algorithms is evaluated via simulation of a generic maneuvering target tracking problem.

This study presents an adaptation of a novel estimation methodology to the general nonlinear three-dimensional problem of tracking a maneuvering target. The two-step optimal estimator (TSE) suggests an attractive alternative to the standard extended Kalman filter (EKF). A superior performance is accomplished by dividing the estimation problem into two steps: a linear first step and a nonlinear second step. The target tracking performance of the TSE is shown to be better than an EKF implemented in either inertial or modified spherical coordinates. In the passive case, where bearing/elevation angles only are measured, the TSE yields excellent range and target acceleration estimates. In the active case, where range measurement is available as well, a homing missile employing closed-loop optimal guidance based on the TSE state estimates obtains smaller miss distances than with either versions of the EKF.

A method for calculating the Euclidean distance in higher order spectra feature space is proposed. The results show that the feature vector extracted from the average profile in a small target-aspect sector has better generalization performance than the average feature vector in the same sector when both of them are used as feature templates in high resolution range profile (HRRP)-based radar automatic target recognition (RATR).

The aim of the present study is to investigate interactions between vision and audition during a target acquisition task performed in a virtual environment. We measured the time taken to locate a visual target (acquisition time) signalled by auditory and/or visual cues in conditions of variable visual load. Visual load was increased by introducing a secondary visual task. The auditory cue was constructed using virtual three-dimensional (3D) sound techniques. The visual cue was constructed in the form of a 3D updating arrow. The results suggested that both auditory and visual cues reduced acquisition time as compared to an uncued condition. Whereas the visual cue elicited faster acquisition time than the auditory cue, the combination of the two cues produced the fastest acquisition time. The introduction of secondary visual task differentially affected acquisition time depending on cue modality. In conditions of high visual load, acquiring a target signalled by the auditory cue led to slower and more error-prone performance than acquiring a target signalled by either the visual cue alone or by both the visual and auditory cues. Keywords: Target acquisition; Visual load; Dual task; Virtual reality; Head related transfer functions; Three-dimensional sound; Multisensory integration

One of the central problems in Automated Target Recognition is to accommodate the infinite variety of clutter in real military environments. The principle focus of our paper is on the construction of metric spaces where the metric measures the distance between objects of interest invariant to the infinite variety of clutter. Such metrics are formulated using second-order random field models. Our results indicate that this approach significantly improves detection/ classification rates of targets in clutter. Index Terms--Riemannian metrics, deformable templates, Automated Target Recognition (ATR).

At what stage does semantic priming affect accuracy in target search? In two experiments, participants viewed two streams of stimuli, each including a target word among distractors. Stimulus onset asynchronies (SOAs) between the targets (T1 and T2) ranged from 53 to 213 msec. A word semantically related to one or neither of the targets preceded each trial. In Experiment 1, participants were instructed to report both targets. Although more primed than unprimed targets were reported, there was no cost for unprimed words. A strong interaction between SOA and T1 versus T2 was found, but priming did not interact with either variable. In Experiment 2, only related targets were reported. Performance was similar to that for primed targets in Experiment 1. Semantic priming does not seem to modulate how attentional resources are initially allocated between targets, but instead affects a later stage of processing, the point at which a target word reaches lexical identification.

Automatic target recognition in uncontrolled conditions is a difficult task because many parameters are involved. This study deals with the recognition of targets under limited out-of-plane rotations while maintaining invariance to ambient light illumination. Contrast invariance is achieved by using the recently developed locally adaptive contrast-invariant filter, a method that yields correlation peaks whose values are invariant under any linear transformation of intensity. To reduce the sensitivity to the orientation of the object we replace the reference in the nonlinear filter by a synthetic discriminant filter. The range used for out-of-plane rotations was 40 degrees with a depression angle of 20 degrees. We present results for unsegmented targets on complex backgrounds with the presence of false targets. OCIS codes: 070.0070, 070.4550, 100.2000, 100.5010, 150.2950.

The batch Maximum Likelihood Estimator, combined with the Probabilistic Data Association Algorithm (ML-PDA), has been shown to be effective in acquiring low observable (LO)--low signal-to-noise ratio (SNR)--nonmaneuvering targets in the presence of heavy clutter. The use of signal strength or amplitude information (AI) in the ML-PDA estimator facilitates the acquisition of weak targets. We present an adaptive algorithm, which uses the ML-PDA estimator with A1 in a sliding-window fashion, to detect possibly maneuvering targets in heavy clutter using electro-optical (EO) sensors. The initial time and the length of the sliding window are adjusted adaptively according to the information content of the received measurements. A track validation scheme via hypothesis testing is developed to confirm the estimated track, that is, the presence of a target, in each window. The sliding-window ML-PDA approach, together with track validation, enables early track detection by rejecting noninformative scans, target reacquisition in case of temporary target disappearance, and the handling of targets with velocities evolving over time. We demonstrate the operation of the adaptive sliding-window ML-PDA estimator on a real scenario for tracking a fast-moving F1 Mirage fighter jet using an imaging sensor. The proposed algorithm is shown to detect the target, which is hidden in as many as 600 false alarms per scan, 10 frames earlier than the Multiple Hypothesis Tracking (MHT) algorithm. This ability to successfully process large amounts of data, with near real-time performance, under time-varying low SNR conditions makes the proposed estimator superior to other existing approaches.

Two algorithms are derived for the problem of tracking a manoeuvring target based on a sequence of noisy measurements of the state. Manoeuvres are modeled as unknown input (acceleration) terms entering linearly into the state equation and chosen from a discrete set. The expectation maximization (EM) algorithm is first applied, resulting in a multi-pass estimator of the MAP sequence of inputs. The expectation step for each pass involves computation of state estimates in a bank of Kalman smoothers tuned to the possible manoeuvre sequences. The maximization computation is efficiently implemented using the Viterbi algorithm. A second, recursive estimator is then derived using a modified EM-type cost function. To obtain a dynamic programming recursion, the target state is assumed to satisfy a Markov property with respect to the manoeuvre sequence. This results in a recursive but suboptimal estimator implementable on a Viterbi trellis. The transition costs of the latter algorithm, which depend on filtered estimates of the state, are compared with the costs arising in a Viterbi-based manoeuvre estimator due to Averbuch, et al. (IEEE Transactions on Aerospace and Electronic Sytstme, 27, 3 (1991), 550-563). It is shown that the two criteria differ only in the weighting matrix of the quadratic part of the cost function. Simulations are provided to demonstrate the performance of both the batch and recursive estimators compared with Averbuch's method and the interacting multiple model (IMM) filter.

It is difficult to identify the mode of operation of a hybrid system, while at the same time tracking the state, without a direct modal measurement. It is shown that, for one estimation algorithm, modal identification can be improved with an adjustment of the measurement update.

Moving target detection and imaging results for an X band spotlight synthetic aperture radar (SAR) system that utilizes an along-track monopulse configuration for its data collection is presented. The theoretical foundation of the processing that is used on these data is based on our earlier work in which a two-dimensional signal subspace processing (adaptive filtering) method was used to calibrate the monostatic and bistatic radars of the monopulse SAR system. The blind calibration of the two channels enables the user to null the stationary scene, and detect the moving targets. Next, a measure that we call SAR ambiguity function is used to estimate the relative speed of a detected moving target. The resultant estimate is then used to image the moving target.

Tracking of a group of targets temporarily operating in formation is relevant for military air-surveillance. An optimal sequential multiple hypothesis tracking (MHT) track extraction and track maintenance method for well-separated single targets embedded in clutter has been described in an earlier paper. However, applied to a group of closely spaced targets, its performance is not acceptable due to data miscorrelation. Here the approach has been generalized to include extraction and maintenance of a maneuvering formation in the presence of clutter. The unknown number of targets and imperfect detection when the group newly appears is modeled within the MHT framework. The generalized method solves the problem very well and can also track a formation after the targets separate.

High resolution range profiles usually suffer from range sidelobe artifacts which cause reduction in the dynamic range. The sidelobes can be greatly reduced by a deconvolution technique called Coherent CLEAN. The Coherent CLEAN algorithm is based on the assumption that the scene consists of isolated and independent targets. However, many real-life targets are contiguous. Even if we approximate the contiguous targets by very closely spaced point sources, they can hardly be assumed to radiate independently. The sidelobes and the mainlobes of these closely spaced point sources can interact constructively and destructively causing spurious peaks and peak mislocations. These problems are studied and a variation in the existing Coherent CLEAN algorithm, called Sequence CLEAN, is proposed. Sequence CLEAN is found to work well with actual targets.

We present a new method for automatic target/object classification by using the optimum polametric radar signatures of the targets/objects of interest. The state-of-the-art in radar target recognition is based mostly either on the use of single polarimetric pairs or on the four preset pairs of orthogonal polarimetric signatures. Due to these limitations polarimetric radar processing has been fruitful only in the area of noise suppression and target detection. The use of target separability criteria for the optimal selection of radar signal state of polarizations is addressed here. The polarization scattering matrix is used for the derivation of target signatures at arbitrary transmit and receive polarization states (arbitrary polarization inclination angles and ellipticity angles). Then an optimization criterion that minimizes the within class distance and maximizes the between class metrics is used for the derivation of optimum sets of polarimetric states. The results of the application of this approach on real synthetic aperture radar (SAR) data of military vehicles are obtained. The results show that noticeable improvements in target separability and consequently target classification can be achieved by the use of the optimum over non-optimum signatures.

A maximum a-posteriori (MAP) estimator, for multiple target tracks that operate directly from array data, is proposed, where the number of contacts is assumed to be known. The estimator is an iterative algorithm that relies on a nonlinear programming (NLP) penalty method, together with an expectation-maximization (EM) algorithm.

A concept of target perceivability is proposed, related to concepts such as target existence and target observability. A basis is thereby provided for an integrated approach to track initiation, confirmation, termination, and refinement of track maintenance algorithms.

The problem of nonlinear estimation is discussed, and a semi-parametric representation called the Biscay distribution is proposed. The Biscay distribution is combined with the extended Kalman filter (EKF), and a filtering paradigm called the Biscay distribution filter (BDF) is described.

We develop a wavelet denoising scheme to aid an automatic target recognition (ATR) system in recognizing aircraft from high range resolution radar (HRR) signatures. A template matching classification technique is used with templates formed from synthetically generated signatures. The goal of the classification system is to achieve classification accuracy equivalent to that obtained with measured HRR signatures. Results suggest that a large portion of HRR signature content is nondiscriminatory. The wavelet denoising process removes the nondiscriminatory information, thereby leading to remarkable increases in classification accuracy. Results are shown for HRR signatures from six aircraft.

The detection and identification of targets obscured by foliage have been topics of great interest. Several synthetic aperture radar (SAR) experiments have demonstrated promising images of terrain and man-made objects obscured by dense foliage, by using either linear frequency modulation (LFM) or step-frequency waveforms. We present here the methodology and results of a comparative study on foliage penetration (FOPEN) SAR imaging using ultrawideband (UWB) step-frequency and random noise waveforms. A statistical-physical foliage transmission model is developed for simulation applications. The foliage obscuring pattern is analyzed by means of the technique of paired echoes. The results of the comparative study demonstrates the ability of a UHF band UWB random noise radar to be used as a FOPEN SAR. Advantages of the random noise radar system include covert detection and immunity to radio frequency interference (RFI).

This study considers the clutter suppression and feature extraction of multiple moving targets for airborne high range resolution (HRR) phased array radar. To avoid the range migration problems that occur in the HRR radar data, we divide each HRR profile into nonoverlapping low range resolution (LRR) segments. No information is lost due to the division and hence no loss of resolution occurs. We show how to use a vector auto-regressive (VAR) filtering technique to suppress the clutter. Then a relaxation-based parameter estimation algorithm is presented for multiple moving target feature extraction. Numerical results are given to demonstrate the effectiveness of the algorithm.

A new 1-D template-based automatic target recognition (ATR) algorithm is developed and tested on high range resolution (HRR) profiles formed from synthetic aperture radar (SAR) images of targets taken from the Moving and Stationary Target Acquisition and Recognition (MSTAR) data set. In this work, a superresolution technique known as High Definition Vector Imaging (HDVI) is applied to the HRR profiles before the profiles are passed through ATR classification. The new 1-D ATR system using HDVI demonstrates significantly improved target recognition compared with previous 1-D ATR systems that use conventional image processing techniques. This improvement in target recognition is quantified by improvement in probability of correct classification (PCC). More importantly, the application of HDVI to HRR profiles helps to maintain the same ATR performance with reduced radar resource requirements.

The derivation of a completely adaptive polarimetric coherent scheme to detect a radar target against a Gaussian background is presented. A previously proposed Generalized Likelihood Ratio Test (GLRT) polarimetric detector is extended to the case of a general number of channels; this exploits the polarimetric characteristics of the received radar echoes to improve the detection performance. Together with the fully adaptive scheme, a model-based detector is derived that has a lower estimation loss. A complete theoretical expression is derived for the detection performance of both proposed polarimetric detectors. They are shown to have Constant False Alarm Rate (CFAR) when operating against Gaussian clutter, but to be sensitive to deviations from the Gaussian statistic (problem solved in Part H of this work). The application to recorded radar data demonstrates the performance improvement achievable in practice.

A target detection problem in radar imaging is examined, for which the covariance matrix of unknown Gaussian clutter has block diagonal structure. Adaptive detection algorithms are designed, using both the generalized likelihood ratio (GLR) and the invariance principles.

********** This experiment seeks to examine the relationships among three advanced technology features (presentation of target cuing, reliability of target cuing, and level of image reality and the attention and […], Michelle Yeh received her Ph.D. in engineering psychology from the University of Illinois at Urbana-Champaign in 2000. She is a member of the D560 HCI and Visualization Group at MITRE in Bedford, Massachusetts. Christopher D. Wickens is a professor of experimental psychology, head of the Aviation Research Laboratory, and associate director of the Institute of Aviation at the University of Illinois at Urbana-Champaign. He also holds appointments in the Department of Mechanical and Industrial Engineering and the Beckman Institute of Science and Technology. He received his Ph.D. in psychology from the University of Michigan in 1974.

Using selected bispectra with maximum interclass separability as feature vectors of signals is proposed. The purpose is to avoid problems involving redundant and/or harmful bispectra in signal classifications.

Two detection schemes for the detection of a spatially distributed, Doppler-shifted target in non-Gaussian clutter are developed. The non-Gaussian clutter is modeled as a spherically invariant random vector (SIRV) distribution. For the first detector, called the non-scatterer density dependent generalized likelihood ratio test (NSDD-GLRT), the detector takes the form of a sum of logarithms of identical functions of data from each individual range cell. It is shown under the clutter only hypothesis, that the detection statistic has the chi-square distribution so that the detector threshold is easily calculated for a given probability of false alarm [P.sub.F]. The detection probability [P.sub.D] is shown to be only a function of the signal-to-clutter power ratio [(S/C).sub.opt] of the matched filter, the number of pulses N, the number of target range resolution cells J, the spikiness of the clutter determined by a parameter of an assumed underlying mixing distribution, and [P.sub.F]. For representative examples, it is shown that as N, J, or the clutter spikiness increases, detection performance improves. A second detector is developed which incorporates a priori knowledge of the spatial scatterer density. This detector is called the scatterer density dependent GLRT (SDD-GLRT) and is shown for a representative case to improve significantly the detection performance of a sparsely distributed target relative to the performance of the NSDD-GLRT and to be robust for a moderate mismatch of the expected number of scatterers. For both the NSDD-GLRT and SDD-GLRT, the detectors have the constant false-alarm rate (CFAR) property that [P.sub.F] is independent of the underlying mixing distribution of the clutter, the clutter covariance matrix, and the steering vector of the desired signal.

The classification of ship targets using low resolution down-range radar profiles together with preprocessing and neural networks is investigated. An implementation of the Fourier-modified discrete Mellin transform is used as a means for extracting features which are insensitive to the aspect angle of the radar. Kohonen's self-organizing map with learning vector quantization (LVQ) is used for the classification of these feature vectors. The use of a feed-forward network trained with the back-propagation algorithm is also investigated. The classification system is applied to both simulated and real data sets. Classification accuracies of up to 90% are reported for the real data, provided target aspect angle information is available to within an error not exceeding 30 deg.

In 1981 Ball and Sekuler showed that a briefly flashed line which primed the subject to the direction of target motion improved the detection of a target. Our aim was to study whether a pointing arrow and random-dot patterns that move in one of four possible directions can prime detection of motion for higher-contrast stimuli. When the motion of a target was primed and the cue validity was 100%, the target's position was more easily detected. Improvement was significant when the direction indicated by the cue and a target's direction were either the same or opposite relative to each other. When the subjects (n = 5 each experiment) did not know the direction of the target before the presentation of the stimulus field, no improvement was found. A discussion of attention to the cue is presented.

Research was conducted to examine a variable-structure multiple model state estimation method used for maneuvering aircraft radar tracking. The switching grid interacting multiple model (SGIMM) and adaptive grid (AG) IMM algorithms and the conventional fixed grid (FGIMM) filter were evaluated and compared over various flight scenarios. Results indicate that the AGIMM filter offers a better performance than the SGIMM while both of them offer better tracking properties than FGIMM.

This communication reports an analysis of Tn5/IS50 target site selection by using an extensive collection of Tn5 and IS50 insertions in two relatively small regions of DNA (less than 1 kb each). For both regions data were collected resulting from in vitro and in vivo transposition events. Since the data sets are consistent and transposase was the only protein present in vitro, this demonstrates that target selection is a property of only transposase. There appear to be two factors governing target selection. A target consensus sequence, which presumably reflects the target selection of individual pairs of Tn5/IS50 bound transposase protomers, was deduced by analyzing all insertion sites. The consensus Tn5/IS50 target site is A-GNTYWRANC-T. However, we observed that independent insertion sites tend to form groups of closely located insertions (clusters), and insertions very often were spaced in a 5-bp periodic fashion. This suggests that Tn5/IS50 target selection is facilitated by more than two transposase protomers binding to the DNA, and, thus, for a site to be a good target, the overlapping neighboring DNA should be a good target, too. Synthetic target sequences were designed and used to test and confirm this model.

Two types of modular neural networks exhibited optimal results for automatic target recognizer classification. Network complexity and classification results were better when input features were decomposed compared to a fully connected network. Similar performance results were obtained in the application of data decomposition through individual neural networks, as each expert network faces a simpler classification problem.

Linear and nonlinear adaptive networks are used for target discrimination in synthetic aperture radar imagery. Problems associated with the two-parameter constant false alarm rate detector led to the use of the quadratic gamma discriminator (QGD) as a classifier. This classifier is further improved in terms of linear processing as a multilayer perceptron (MLP) topology called the nonlinear QGD (NL-QGD).

A simultaneous top-down-and-bottom-up concurrent matching procedure eliminates the disadvantages of top-down and backtracking methods for foveal automatic target recognition through a multiresolution neural network. The multilayer Hopfield neural network allows greater confirmation accuracy of potential matches across different resolution levels. The method employs gaze control, which allows intelligent resource allocation for the foveal images.

Observers viewed an animated ascending or descending target that varied in size and velocity across trials and appeared either (a) in isolation, (b) to slide along one side of a single larger stationary object, or (c) to slide between two larger stationary objects. Targets vanished without warning, and displacements (i.e., differences between actual and remembered final position) along the axis of motion and orthogonal axis were measured. Forward displacement (a) decreased with increases in implied friction, (b) increased with increases in target size for descending targets, and (c) decreased with increases in target size for ascending targets. When a larger stationary object was to one side of the target, orthogonal displacement was toward that object; when no object or objects on both sides were present, orthogonal displacement was near zero. Results are consistent with previous findings and speculation on the effects of representational friction, memory averaging, and target size on memory.

Sixteen subjects with normal or corrected-to-normal vision participated in a a study which examined the degree to which a cursor-catching feedback using dynamic cursor gain can lessen target acquisition times in contrast to visual only and combined visual-tactual feedback. Each subject completed four target feedback conditions, namely, visual only, visual-tactual, visual-dynamic and combined-dynamic, which were tested under three outside-target gain levels. Results showed that a tactually felt 'hole' field in the target area gives a stronger 'catching effect' sense than a dynamic target gain.