When there exists the limitation of communication bandwidth between sensors and a fusion center, one needs to optimally precompress sensor outputs-sensor observations or estimates before sensors' transmission in order obtain constrained optimal estimation at center terms linear minimum error variance criterion, when an allowed performance loss constraint exists, design dimension data. This paper will answer above questions by using matrix decomposition, pseudo-inverse, eigenvalue techniques.

Distributed estimation fusion is concerned with the combination of local estimates from multiple distributed sensors to produce a fused result. In this paper, we characterize as posterior probability densities, and assume that they all belong parametric family. Our starting point consider family Riemannian manifold by introducing Fisher information metric. From perspective geometry, density formulated an informative barycenter in space densities sought minimizing sum its squared geodesic...

Under the assumption of independent observation noises across sensors, Bar-Shalom and Campo proposed a distributed fusion formula for two-sensor systems, whose main calculation is inverse submatrices error covariance two local estimates instead itself. However, corresponding simple estimation absent in general multisensor system. In this paper, an efficient iterative algorithm without any restrictive on noise (i.e., sensors system, direct computation Moore-Penrose generalized joint are not...

In this paper, an extension of the standard Kalman filtering for dynamical systems with white noises to finite-time correlated is addressed. Although one can augment state vector in a time-variant moving average process which models noise, and then use obtain optimal estimate mean square error sense, direct recursion original general cases was pursued owing lower computational complexity. By decomposing gain two recursively represented factors increasing some recursive terms (for more than...

This paper deals with adaptive radar detection of targets in the presence Gaussian disturbance sharing a block-diagonal covariance structure. The problem is formulated according to very general signal model, which contains point-like, range-spread, and subspace target (or targets) as special instances. Hence, unified study on resulting handled use invariance theory. obtained results, including an appropriate transformation group, maximal invariant induced invariant, are proven consistent...

The invariance principle is adopted to develop an exhaustive study for adaptive detection of range-spread targets in Gaussian noise sharing a block-diagonal covariance structure. For this problem, the usual generalized likelihood ratio intractable. In paper, we first determine largest group affine transformations that does not alter decision problem. Then, maximal invariant identified by derived, which can characterize totality detectors and extends existing results point-target case. A...

In this paper, we present an order-recursive formula for the pseudoinverse of a matrix. It is variant well-known Greville [SIAM Rev., 2 (1960), pp. 578--619] formula. Three forms proposed are presented three different matrix structures. Compared with original formula, formulas have certain merits. For example, they reduce storage requirements at each recursion by almost half; more convenient deriving recursive solutions optimization problems involving pseudoinverses. Regarding applications,...

This paper addresses the estimation of large-dimensional covariance matrices under both normal and nonnormal distributions. The shrinkage estimators are constructed by convexly combining sample matrix a structured target matrix. optimal oracle intensity is obtained analytically for any prespecified in class which includes various such as banding, thresholding, diagonal, block diagonal matrices. After deriving unbiased consistent estimates some quantities involving unknown population matrix,...

This letter deals with similarity parameter selection for knowledge-aided covariance matrix estimation in adaptive radar signal processing. Starting from the observation that maximum likelihood estimate of interference under a constraint admits closed-form expression, which depends on parameter, an procedure is devised to get free estimator. The technique based expected principle and requires solution implicit equation, can be efficiently pursued via bisection method due monotonicity...

This study investigates the design of an optimal linear estimator for a class discrete‐time systems with correlated stochastic parameter matrices and noises. The considered are endowed following two main features: (i) cross‐correlated involved in state observation equations assumed (ii) process noises have cross‐correlation at same time instant. A decorrelation framework is established to reconstruct such systems. With equivalent transformation original dynamic resulting from decorrelating...

This paper studies linear distributed estimation of an unknown random parameter vector in a bandwidth-constrained multisensor network. To meet the bandwidth limitations, each sensor converts its observation into low-dimensional datum via suitable transformation. Then, fusion center estimates by linearly combining all received data, aiming at minimizing mean square error. The main purpose this is to jointly determine compression dimension (referred as assignment) and design corresponding...

This paper studies the problem of detecting range-spread targets in (possibly non-Gaussian) clutter whose joint distribution belongs to a very general family complex matrix-variate elliptically contoured distributions. Within family, we explore invariance with respect both distributional type and relevant parameters. Several groups are used describe these mechanisms, relationship is revealed between group constant false alarm rate (CFAR) properties terms model parameters, generator function,...

Graph embedding (GE) provides an effective way to reveal the intrinsic feature of high-dimensional data on foundation preserving topological properties. Under framework GE, hyperspectral image can be represented by a weighted graph, where pixels and similarities among them are treated as vertices edge weights, respectively. In this article, adaptive reference-related GE (ARGE) method is proposed efficaciously obtain low-dimensional improve computational efficiency. The ARGE composed two...

This paper presents a novel composite heuristic algorithm for global optimization by organically integrating the merits of water cycle (WCA) and gravitational search (GSA). To effectively reinforce exploration exploitation algorithms reasonably achieve their balance, modified WCA is first put forward to strengthen its performance introducing concept basin, where position solution also considered into assignment sea or river streams, number guider solutions adaptively reduced during process....

In this paper, the communication direction problem of a two-sensor tandem binary decision system is considered. Rigorous analysis shows that performance from sensor with higher noise power to lower not always better than reverse when signal and noises are both Gaussian. This result can be extended more general without assumption specific data distribution. seems somewhat counterintuitive but has significance for optimization design direction. Computer experiments support our analytic results...

The problem of state estimation for discrete-time stochastic time-varying systems in the presence unknown process inputs or disturbances is addressed this paper. A Kalman-type filter proposed, and optimal oracle gain sense minimizing mean squared error estimate obtained. To tackle quantities matrix, a nonlinear equation introduced its solution taken as inputs, then, novel equation-based input filtering (NEUIF) proposed. scalar-based iterative algorithm related fixed point developed so that...