Amy R. Reibman

Optimal Detection and Performance of Distributed Sensor Systems

Global optimization of a distributed sensor detection system withfusion is considered, where the fusion rule and local detectors aresolved to obtain overall optimal performance. This yields coupledequations for the local detectors and the fusion center.The detection performance of the distributed system with fusionis developed. The globally optimal system performance is comparedwith two suboptimal systems. Receiver operating characteristics(ROCs) are computed numerically for the problem of detecting aknown signal embedded in non-Gaussian noise.

Redundancy rate-distortion analysis of multiple description coding using pairwise correlating transforms

The objective of multiple description coding (MDC) is to encode a source into two (or more) bitstreams supporting two quality levels of decoding. A high-quality reconstruction should be decodable from the two bitstreams together, while lower, but still acceptable, quality reconstructions should be decodable from either of the two individual bitstreams. This paper describes techniques for meeting MDC objectives in the framework of standard transform-based image coding through the design of pairwise transforms.

Design and Performance Comparison of Distributed Detection Networks

Optimization of a distributed detection network using theminimum global cost criterion results in local processors thatindividually form the likelihood ratio when the input observationvectors are statistically independent. In addition, the localthresholds and the network performance can be expressed as afunction of the receiver operating characteristics (ROCs) of the localprocessors. The performance of rive distributed networks arecompared numerically using local ROCs from the conic ROCfamily.

DCT-based embedded coding for packet video☆

Abstract We present an image compression algorithm that is useful for transmitting video over packet networks. The algorithm modifies those in references [4–6, 8, 10] by incorporating motion-compensated prediction. The low and high priority images can be computed in parallel, and the algorithm provides improved motion-compensated prediction of low-frequency discrete cosine transform (DCT) coefficients.

Hough transform and signal detection theory performance for images with additive noise

Abstract The line detection performance and sensitivity to the noise distribution of the Hough transform and two signal detection theory processors are evaluated quantitatively (using receiver operating characteristics (ROC)) and compared for images corrupted by each of several types of additive noise. The types of noise distributions considered are Gaussian, uniform, and Laplacian. The two types of signal detection theory processors considered are the optimal detector for additive, Gaussian noise and the optimal detector for additive, Laplacian noise. The performances for these noise distributions are interesting to compare because they vary widely in the thickness of the tails of their probability density functions. The Gaussian processor and the Hough transform are found to be much less sensitive to noise type than the Laplacian processor.

Optimal design and performance of distributed signal detection systems with faults

The design and performance of distributed signal detection systems with processor failures is presented. Using a general processor fault model, which also models an imperfect channel, the optimal design given prior failure probabilities is considered. The optimal design provides a performance bound for each system when failures may occur. It is shown that for general distributed systems, a likelihood ratio test is the optimal design for each local processor, provided failures are independent of the received observation vectors. The optimal design performs significantly better than the design that assumes no failures. It is shown that the fusion network performs better than a single centralized processor for many channels. This illustrates that, for the fusion network, the performance gained by decreasing the effect of failures outweighs the performance lost by distributing the processing. The serial network always performs worse than centralized; its performance is bounded as the number of channels increases. >

On determining the design of fusion detection networks

Methods are explored to determine an efficient design procedure for large, distributed fusion networks. An efficient design will not require optimization over the discrete fusion rule. The authors present several approaches to developing efficient designs, assuming equal local thresholds. However, the designs developed are not necessarily optimal, given this assumption. The authors also present the numerical detection performance (probability of error) produced from optimizing each performance measure. In addition, they present the conditions under which the obvious design (using the local minimum-probability-of-error criterion) is not optimal.<<ETX>>

Asymptotic quantization for signal estimation with noisy channels

The authors consider signal estimation using a communication system which consists of a quantizer, a noisy channel, and a decision device. By assuming that the lengths of the quantization intervals become arbitrarily small, they derive the asymptotic signal estimation error for the communication system. The asymptotic error consists of an interaction term which depends on both the quantization error and the channel error. The most significant property of the interaction term is that it can be negative. Next, the authors use the companding approach to find the optimal uniform quantizer. They then suggest using a suboptimal piecewise uniform quantizer to approximate the optimal nonuniform quantizer. Finally, they present several analytic and numerical results to justify the validity of the obtained asymptotic expression.<<ETX>>

Source coding for decentralized estimation systems

Parameter estimation in decentralized systems with distributed processors is considered. The authors restrict the local processors to be quantizers and consider the optimal design of the system to minimize the estimation error. They derive the necessary conditions of the optimal system based on different distortion functions. In particular, they show that using Fishers information as a distortion function can simplify the design of quantizers. An expression for the asymptotic quantization error of scalar quantizers is also derived. Numerical results show that using Fishers information can simplify the design and achieve good performance.<<ETX>>

Performance of redundant and distributed detection systems with processor faults

Examines the performance of signal detectors when processors can have faults. The fault tolerance of two systems, the redundant system and the distributed fusion network, are examined. The optimal fault-tolerant design for each system is presented. The performance of the fault-tolerant systems is compared to the faulty simplex processor. In the absence of processor failures, the redundant system is identical to the simplex processor. However, when failures occur, the redundant system performs better than the simplex processor. The fault-free fusion network does not perform as well as the fault-free simplex processor. However, in the cases examined, when failures occur in systems with many (N> 12) channels, the fusion network performs better than the simplex processor.<<ETX>>