Optimal distributed decision fusion

The problem of decision fusion in distributed sensor systems is considered. Distributed sensors pass their decisions about the same hypothesis to a fusion center that combines them into a final decision. Assuming that the sensor decisions are independent of each other for each hypothesis, the authors provide a general proof that the optimal decision scheme that maximizes the probability of detection at the fusion for fixed false alarm probability consists of a Neyman-Pearson test (or a randomized N-P test) at the fusion and likelihood-ratio tests at the sensors     

Optimum multisensor fusion of correlated local decisions

A distributed detection system consisting of a number of local detectors and a fusion center is considered. Each detector makes a decision for the underlying binary hypothesis testing problem based on its own observation and transmits its decision to the fusion center where the global decision is derived. The local decision rules are assumed to be given, but the local decisions are correlated. The correlation is generally characterized by a finite number of conditional probabilities. The optimum decision fusion rule in the Neyman-Pearson sense is derived and analyzed. The performance of the distributed detection system versus the degree of correlation between the local decisions is analyzed for a correlation structure that can be indexed by a single parameter. System performance as well as the performance advantage of using a larger number of local detectors degrade as the degree of correlation between local decisions increases     

Weak Signal Detection in Non-Gaussian Noise of Unknown Level

An adaptive threshold detector to test for the presence of a weak signal in additive non-Gaussian noise of unknown level is discussed. The detector consists of a locally optimum detector, a noise level estimator, and a decision device. The detection threshold is made adaptive according to the information provided by the noise level estimator in order to keep a fixed false-alarm probability. Asymptotic performance characteristics are obtained indicating relationships among the basic system parameters such as the reference noise sample size and the underlying noise statistics. It is shown that, as the reference noise sample size is made sufficiently large, the adaptive threshold detector attains the performance of a corresponding locally optimum detector for detecting the weak signal were the noise level known.     

A Suboptimum Rank Test for Nonparametric Radar Detection

The optimum rank detector structure, in the Neyman-Pearson sense and under Gaussian noise conditions, is approximated by a suboptimum structure that depends on an adjustable parameter. This new rank detector, which operates on radar video signal, includes other well-known detectors as particular cases. The asymptotic relative efficiency (ARE) of the proposed rank detector is computed, with its maximum value the ARE of the locally optimum rank detector (LORD). The detection probability versus signal-to-noise ratio, and the effects of interfering targets are also calculated by Monte-Carlo simulations for different parameter values.     

Asymptotic Performance of Sparse Signal Detection Using Convex Programming Method

The detection of sparse signals against background noise is considered. Detecting signals of such kind is difficult since only a small portion of the signal carries information. Prior knowledge is usually assumed to ease detection. In this paper, we consider the general unknown and arbitrary sparse signal detection problem when no prior knowledge is available. Under a Neyman-Pearson hypothesis-testing framework, a new detection scheme is proposed by combining a generalized likelihood ratio test (GLRT)-like test statistic and convex programming methods which directly exploit sparsity in an underdetermined system of linear equations. We characterize large sample behavior of the proposed method by analyzing its asymptotic performance. Specifically, we give the condition for the Chernoff-consistent detection which shows that the proposed method is very sensitive to the 2 norm energy of the sparse signals. Both the false alarm rate and the miss rate tend to zero at vanishing signal-to-noise ratio (SNR), as long as the signal energy grows at least logarithmically with the problem dimension. Next we give a large deviation analysis to characterize the error exponent for the Neyman-Pearson detection. We derive the oracle error exponent assuming signal knowledge. Then we explicitly derive the error exponent of the proposed scheme and compare it with the oracle exponent. We complement our study with numerical experiments, showing that the proposed method performs in the vicinity of the likelihood ratio test (LRT) method in the finite sample scenario and the error probability degrades exponentially with the number of observations.     

Optimization of multiframe target detection schemes

We optimize the performance of multiframe target detection (MFTD) schemes under extended Neyman-Pearson (NP) criteria. Beyond the per-track detection performance for a specific target path in conventional MFTD studies, we optimize the overall detection performance which is averaged over all the potential target paths. It is shown that the overall MFTD performance is limited by the mobility of a target and also that optimality of MFTD performance depends on how fully one ran exploit the information about the target dynamics. We assume a single target situation and then present systematic optimization by formulating the MFTD problems as binary composite hypotheses testing problems. The resulting optimal solutions suggest computationally efficient implementation algorithms which are similar to the Viterbi algorithm for trellis search. The optimal performances for some typical types of target dynamics are evaluated via Monte-Carlo simulation     

Optimal serial distributed decision fusion

The problem of distributed detection involving N sensors is considered. The configuration of sensors is serial in the sense that the Jth sensor decides using the decision it receives along with its own observation. When each sensor uses the Neyman-Pearson test, the probability of detection is maximized for a given probability of false alarm, at the Nth stage. With two sensors, the serial scheme has a performance better than or equal to the parallel fusion scheme analyzed in the literature. Numerical examples illustrate the global optimization by the selection of operating thresholds at the sensors     

Numerical Results on the Convergence of Relative Efficiencies

This paper considers the detection of a known constant signal in an additive non-Gaussian noise under the assumptions of discrete time and statistically independent noise samples. The objective is to determine how large sample size must be before the easily computed asymptotic relative efficiency becomes a valid measure of performance. The exact small-sample error probabilities are calculated for a Neyman-Pearson optimal nonlinear detector consisting of a zeromemory nonlinearity followed by summation and threshold comparison. "Large-tailed" noise having a double exponential distribution is used as an example. The exact distribution of the test statistics for a linear detector and for the Neyman-Pearson optimal detector are calculated. Then the relative efficiency of the Neyman-Pearson optimal detector, as compared to a linear detector, is computed in order to study the rate of approach of the relative efficiency to its asymptotic value.     

A unified approach to the design of decentralized detection systems

A unified approach to the design of decentralized detection networks with arbitrary topologies is proposed and analyzed. In this approach, a decentralized detection system of arbitrary topology is represented by a communication matrix that specifies the interconnection structure of the detection network. This matrix is used to derive the general decision rule using the person-by-person-optimal (PBPO) solution methodology. It is shown that the PBPO decision rule is a likelihood ratio test for statistically independent observations. The threshold of the test is shown to be a function of the decision input vector of the detector under consideration. This unified approach is used to obtain the PBPO decision rules of decentralized detection systems with various topologies. Various results in the literature are verified. In addition, the PBPO decision rules for a decentralized detection system with peer communication are presented. Numerical examples are presented for illustration     

Incoherent radar detection in compound-Gaussian clutter

The detection of incoherent pulse trains in compound-Gaussian disturbance with known spectral density is dealt with here. Two alternative approaches are investigated, The first, assuming perfect knowledge of the signal fluctuation law and implementing the Neyman-Pearson test on the observed waveform, turns out to be not applicable to the radar problem. The second, instead, relying on the generalized likelihood ratio optimization strategy, leads to a canonical detector, whose structure is independent of the clutter amplitude probability density function. Interestingly, this detector turns out to be constant false-alarm rate in the sense that threshold setting does not require any knowledge as to the clutter distribution, Moreover, since such a processor is not implementable in real situations, we also present an FFT-based (fast Fourier transform) suboptimum structure. Finally, we give closed-form formulas for the detection performance of both receivers, showing that both of them largely outperform the square-law detector, especially in the presence of very spiky clutter     

Detection in Laplace Noise

The discrete time detection of a known constant signal in white stationary Laplace noise is considered. Exact expressions describing the performance of both the Neyman-Pearson optimal detector and the suboptimal linear detector are presented. Also, graphs of the receiver operating characteristics are given. The actual performance of the Neyman-Pearson optimal detector is compared to that predicted by a Gaussian approximation to the distribution of the test statistic.     

Distributed decision fusion using empirical estimation

The problem of optimal data fusion in multiple detection systems is studied in the case where training examples are available, but no a priori information is available about the probability distributions of errors committed by the individual detectors. Earlier solutions to this problem require some knowledge of the error distributions of the detectors, for example, either in a parametric form or in a closed analytical form. Here we show that, given a sufficiently large training sample, an optimal fusion rule can be implemented with an arbitrary level of confidence. We first consider the classical cases of Bayesian rule and Neyman-Pearson test for a system of independent detectors. Then we show a general result that any test function with a suitable Lipschitz property can be implemented with arbitrary precision, based on a training sample whose size is a function of the Lipschitz constant, number of parameters, and empirical measures. The general case subsumes the cases of nonindependent and correlated detectors.     

Optimum array detector for a weak signal in unknown noise

We study the design of constant false-alarm rate (CFAR) tests for detecting a rank-one signal in the presence of background Gaussian noise with unknown spatial covariance. We look at invariant tests, i.e., those tests whose performance is independent of the nuisance parameters, like the background noise covariance. Such tests are shown to have the desirable CFAR property. We characterize the class of all such tests by showing that any invariant decision statistic can be written as a function of two basic statistics which are in fact the adaptive matched filter (AMF) statistic and Kelly's generalized likelihood ratio statistic. Further, we establish an optimum test in the limit of low signal-to-noise ratio (SNR), the locally most powerful invariant (LMPI) test. We also derive the bound for the probability of detection of any invariant detector, at a fixed false-alarm rate, and compare the LMPI and the published detectors (Kelly and AMF) to it     

Design of multi-sensor attitude determination systems

The design of inexpensive multi-sensor attitude determination systems is discussed. The systems discussed fuse information from a triad of solid state rate gyros with an aiding system mechanized using GPS or magnetometers and accelerometers. Euler angle and quaternion-based sensor fusion algorithms are developed. Methods for gain scheduling and estimator pole placement are presented. Using simulation and flight test results, it is shown that quaternion-based algorithms simplify gain scheduling and improve transient response.     

Bayesian Quickest Signal Detection in a Discrete-Time Observation

This paper deals with the problem of quickest detection of a signal in discrete-time observations where the noise is not necessarily additive. By introducing a new cost function, penalizing the decision delay, in addition to penalizing wrong decisions as in the classical case, a global risk function is derived for use in a Bayesian framework. The minimization of the average risk leads to the optimum Bayesian decision regions, giving the structure of the optimum receiver. Some simplifications for elementary costs and some applications are investigated. The optimum receiver is shown to be a parallel bank of classical optimum filters, each one matched to a particular delay of the signal to be detected. Our approach is shown to apply to the detection of certain changes in a stochastic process.     

CFAR data fusion center with inhomogeneous receivers

Detection systems with distributed sensors and data fusion are increasingly used by surveillance systems. A system formed by N inhomogeneous constant false alarm rate (CFAR) detectors (cell-averaging (CA) and ordered statistic (OS) CFAR detectors) is studied. A recursive formulation of an algorithm that permits a fixed level of false alarms in the data fusion center is presented, to set the optimum individual threshold levels in the CFAR receivers and the optimum `K out of N' decision rule in order to maximize the total probability of detection. The algorithm also considers receivers of different quality or with different communication channel qualities connecting them with the fusion center. This procedure has been applied to several hypothetical networks with distributed CA-CFAR and OS-CFAR receivers and for Rayleigh targets and interference, and it was seen that in general the fusion decision OR rule is not always the best     

Asynchronous distributed detection

Binary parallel distributed-detection architectures employ a bank of local detectors to observe a common volume of surveillance, and form binary local decisions about the existence or nonexistence of a target in that volume. The local decisions are transmitted to a central detector, the data fusion center (DEC), which integrates them to a global target or no target decision. Most studies of distributed-detection systems assume that the local detectors are synchronized. In practice local decisions are made asynchronously and the DFC has to update its global decision continually. In this study the number of local decisions observed by the central detector within any observation period is Poisson distributed. An optimal fusion rule is developed and the sufficient statistic is shown to be a weighted sum of the local decisions collected by the DFC within the observation interval. The weights are functions of the individual local detector performance probabilities (i.e., probabilities of false alarm and detection). In this respect the decision rule is similar to the one developed by Chair and Varshney for the synchronized system. Unlike the Chair-Varshney rule, however, the DFC's decision threshold in the asynchronous system is time varying. Exact expressions and asymptotic approximations are developed for the detection performance with the optimal rule. These expressions allow performance prediction and assessment of tradeoffs in realistic decision fusion architectures which operate over modern communication networks     

Distributed decision fusion in the presence of link failures

The problem of optimal data fusion involving detector unit communication link failures is considered. Two strategies for decision making in presence of link failures are examined and an optimal decision making scheme in the sense of the Neyman-Pearson (N-P) test is proposed. The performance of q+1 reliable links versus q reliable links are examined theoretically, as well as, numerically using the receiver operating characteristics (ROCs)     

A non-contact lie detector using radar vital signs monitor (RVSM)technology

Traditional lie detector testing requires the subject to be physically attached to a variety of sensors. This is impractical for scenarios such as checkpoints where a large number of individuals are entering at a high rate, necessitating the employment of other methods. Currently, checkpoint officers must make a quick decision to determine if an individual is being deceptive, and if, in turn, they should be searched. The remote detection of deception (RDD) concept uses a non-contact sensor to obtain physiological information that can be used to aid the checkpoint officer's decision. Such a device must be able to sense physiological signals from the body that may indicate deception in an unobtrusive and non-contact manner