Large-scale sensor networks, which can monitor the environment at close range with high spatial and temporal resolutions, are expected to play an important role in various applications:
- assessing "health" of machines, aerospace vehicles, and civil-engineering structures,
- environmental, medical, food-safety, and habitat monitoring,
- energy management, inventory control, home and building automation, etc.

Each node will have limited sensing, signal processing, and communication capabilities, but by cooperating with each other they will accomplish tasks that are difficult to perform with conventional centralized sensing systems. Sensor networks are expected to reveal previously unobservable phenomena in the physical world; hence they are currently attracting much attention.

I will present 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. HMRFs belong to the (broader) class of probabilistic graphical models. I will derive a distributed maximum a posteriori (DMAP) method for estimating the hidden random field from the noisy measurements. The DMAP method is computationally simple and applicable to a wide range of sensing environments. It is also localized, implying that the nodes communicate only with their neighbors to obtain the desired results. Localized algorithms are robust to node failures and the communication overhead scales well with increase in the network size.

The proposed approach allows for general measurement and random field models. This is in contrast to previous work on distributed HMRF based signal processing for sensor networks, which focused on developing message passing algorithms for linear Gaussian measurement-error and Markov random field (MRF) process models.

I will apply the proposed approach to the event-region detection problem. Here, my goal will be to detect a region in the environment in which an event of interest has occurred. For example, if the network is capable of sensing concentration of chemical X, then it is of interest to answer the following question: "In which regions is the concentration of chemical X greater than a specified level?" I will consider parametric and nonparametric (empirical likelihood and entropy) measurement-error models and utilize an autologistic MRF process model for event-region detection. I will also demonstrate the performance of the proposed algorithms using numerical simulations.

Finally, I will briefly describe our related work on defect detection for nondestructive evaluation (NDE) of materials, where we applied the HMRF framework to analyze noisy ultrasonic C-scan images of titanium billets.

Aleksandar Dogandzic received the Dipl. Ing. degree (summa cum laude) in Electrical Engineering from the University of Belgrade, Yugoslavia, in 1995, and the M.S. and Ph.D. degrees in electrical engineering and computer science from the University of Illinois at Chicago (UIC) in 1997 and 2001, respectively, under the guidance of Professor Arye Nehorai.

In August 2001, he joined the Department of Electrical and Computer Engineering, Iowa State University, Ames, IA, as an Assistant Professor. His research interests are in statistical signal processing theory and applications.

In 1996 Dr. Dogandzic received the Distinguished Electrical Engineering M.S. Student Award by the Chicago Chapter of the IEEE Communications Society. He was awarded the Aileen S. Andrew Foundation Graduate Fellowship in 1997, the UIC University Fellowship in 2000, and the 2001 Outstanding Thesis Award in the Division of Engineering, Mathematics, and Physical Sciences, UIC. He is the recipient of the 2003 Young Author Best Paper Award and 2004 Signal Processing Magazine Award by the IEEE Signal Processing Society.