Benhong Zhang

Distributed Estimation and Detection for Sensor Networks Using Hidden Markov Random Field Models

We develop a hidden Markov random field (HMRF) framework for distributed signal processing in sensor-network environments. Under this framework, spatially distributed observations collected at the sensors form a noisy realization of an underlying random field that has a simple structure with Markovian dependence. We derive iterated conditional modes (ICM) algorithms for distributed estimation of the hidden random field from the noisy measurements. We consider both parametric and nonparametric measurement-error models. The proposed distributed estimators are computationally simple, applicable to a wide range of sensing environments, and localized, implying that the nodes communicate only with their neighbors to obtain the desired results. We also develop a calibration method for estimating Markov random field model parameters from training data and discuss initialization of the ICM algorithms. The HMRF framework and ICM algorithms are applied to event-region detection. Numerical simulations demonstrate the performance of the proposed approach

Bayesian Complex Amplitude Estimation and Adaptive Matched Filter Detection in Low-Rank Interference

We propose a Bayesian method for complex amplitude estimation in low-rank interference. We assume that the received signal follows the generalized multivariate analysis of variance (GMANOVA) patterned-mean structure and is corrupted by low-rank spatially correlated interference and white noise. An iterated conditional modes (ICM) algorithm is developed for estimating the unknown complex signal amplitudes and interference and noise parameters. We also discuss initialization of the ICM algorithm and propose a (non-Bayesian) adaptive-matched-filter (AMF) signal detector that utilizes the ICM estimation results. Numerical simulations demonstrate the performance of the proposed methods

Nonparametric Probability Density Estimation for Sensor Networks Using Quantized Measurements

We develop a nonparametric method for estimating the probability distribution function (pdf) describing the physical phenomenon measured by a sensor network. The measurements are collected by sensor-processor elements (nodes) deployed in the region of interest; the nodes quantize these measurements and transmit only one bit per observation to a fusion center. We model the measurement pdf as a Gaussian mixture and develop a Fisher scoring algorithm for computing the maximum likelihood (ML) estimates of the unknown mixture probabilities. We also estimate the number of mixture components as well as their means and standard deviation. Numerical simulations demonstrate the performance of the proposed method.

Bayesian NDE Defect Signal Analysis

We develop a hierarchical Bayesian approach for estimating defect signals from noisy measurements and apply it to nondestructive evaluation (NDE) of materials. We propose a parametric model for the shape of the defect region and assume that the defect signals within this region are random with unknown mean and variance. Markov chain Monte Carlo (MCMC) algorithms are derived for simulating from the posterior distributions of the model parameters and defect signals. These algorithms are then utilized to identify potential defect regions and estimate their size and reflectivity parameters. Our approach provides Bayesian confidence regions (credible sets) for the estimated parameters, which are important in NDE applications. We specialize the proposed framework to elliptical defect shape and Gaussian signal and noise models and apply it to experimental ultrasonic C-scan data from an inspection of a cylindrical titanium billet. We also outline a simple classification scheme for separating defects from nondefects using estimated mean signals and areas of the potential defects

Dynamic shadow-power estimation for wireless communications

We present a sequential Bayesian method for dynamic estimation and prediction of local mean (shadow) powers from instantaneous signal powers in composite fading-shadowing wireless communication channels. We adopt a Nakagami-m fading model for the instantaneous signal powers and a first-order autoregressive [AR(1)] model for the shadow process in decibels. The proposed dynamic method approximates predictive shadow-power densities using a Gaussian distribution. We also derive Crame/spl acute/r-Rao bounds (CRBs) for stationary lognormal shadow powers and develop methods for estimating the AR model parameters. Numerical simulations demonstrate the performance of the proposed methods.

Parametric Signal Estimation Using Sensor Networks in the Presence of Node Localization Errors

Signal processing methods developed so far for sensor-network environments have ignored the effects of node localization inaccuracies. We propose a Bayesian framework that accounts for the inherent uncertainties in the node locations (caused by the node localization errors) and develop an estimation method that is robust to these uncertainties. We model the node localization errors as zero-mean Gaussian random vectors whose covariances are known up to a scaling factor. Iterated conditional modes (ICM) and Markov chain Monte Carlo (MCMC) algo- rithms are developed to estimate the signal parameters of interest and construct Bayesian confidence regions for these parameters. The proposed methods are then applied to localize an acous- tic source using energy measurements. Numerical simulations demonstrate the performance of the proposed approach.

Event-Region Estimation for Sensor Networks Under the Poisson Regime

We develop a Bayesian method for event-region estimation in large-scale sensor networks under the Poisson regime. We propose a parametric model for the location and shape of the event region and assume that the unknown signal strength within this region is constant. We adopt a fusion architecture where each node in the network makes a decision locally and then conveys it to a fusion center. Both binary and quantized decisions are considered, corresponding to utilizing one or multiple thresholds (respectively) to make the local decisions. Markov chain Monte Carlo (MCMC) algorithms are derived for simulating from the posterior distributions of the unknown signal, location and shape parameters and for estimating these parameters. Numerical simulations demonstrate the performance of the proposed methods. I. INTRODUCTION

Dynamic power estimation and prediction in composite fading-shadowing channels

We present methods for dynamic estimation and prediction of local mean (shadow) powers from instantaneous signal powers in composite fading-shadowing wireless communication scenarios. We adopt a Nakagami-m fading model for the instantaneous signal powers and a first-order autoregressive - AR(1) - model for the shadow process in decibels. Sequential Bayesian analysis is applied to estimate the shadow powers assuming that the Nakagami-m and AR(1) model parameters are known. We also develop a method for jointly estimating both the shadow powers and unknown model parameters. Numerical simulations demonstrate the performance of the proposed methods.

Complex Signal Amplitude Estimation and Adaptive Detection in Unknown Low-rank Interference

We propose a Bayesian method for complex signal amplitude estimation in low-rank interference. We assume that the received signal follows the generalized multivariate analysis of variance (GMANOVA) patterned-mean structure and is corrupted by low-rank spatially correlated interference and white noise. An iterated conditional modes (ICM) algorithm is developed for estimating the unknown complex signal amplitudes and interference and noise parameters. We also discuss initialization of the ICM algorithm and propose an adaptive-matched- fllter (AMF) signal detector that utilizes the ICM estimation results. Numerical simulations demonstrate the performance of the proposed methods.