|
Ali Pezeshki
Program in Applied and Computational Mathematics,
Princeton University
Abstract: Sensing networks are emerging as a new technology with the potential to enable cost-effective and reliable surveillance. One of the advantages of employing distributed sensors for target detection, compared to an array of closely spaced sensors, is that distributed sensors can provide uncorrelated realizations of the signal, due to the large spatial separation between the sensors. In other words, they can provide spatial diversity. Similar to MIMO communications, where the idea is to exploit independent fadings across spatially separated paths to improve reliability, in distributed sensing the idea is to use spatially separated sensor nodes to measure independent realizations of a signal. However, unlike MIMO communications where spatial diversity is quantified by the notion of diversity gain, a formal measure for spatial diversity in distributed sensing has not been fully developed.
In this talk, we develop a counterpart for distributed sensing of the fundamental tradeoff in wireless communication between rate and reliability. Multiple sensors can cooperate to monitor a single surveillance cell with high fidelity, or they can act independently to monitor multiple cells simultaneously, but with less fidelity. The former is the counterpart of reliability and the latter is the counterpart of rate. We introduce a notion of diversity order for detection with distributed sensors by establishing how network-wide detection error probability behaves as a function of SNR. As a case study, we look at joint Neyman-Pearson (NP) detection and distributed NP detection under a desired false alarm rate (FAR). We show that, in an N-element network, joint detection has FAR-independent diversity order N/r, where r is the number of surveillance cells monitored simultaneously (multiplexing order), while distributed detection has FAR-dependent diversity order. We show that distributed detection has diversity order N/r-k+1, where k is the minimum number of degrees of freedom required to achieve the desired FAR once the detection threshold of local detectors are chosen. We highlight this dependence by giving two examples: one where distributed detection achieves full diversity order N and one where it achieves diversity order one.
This is joint work with Robert Calderbank (Princeton University) and Stephen Howard (DSTO, Australia).
Bio: A. Pezeshki received the BSc and MSc degrees in electrical engineering from the University of Tehran, Tehran, Iran, in 1999 and 2001. He received the PhD degree in electrical engineering from Colorado State University in 2004. During 2005, he was a postdoctoral researcher with the Signal Processing and Communications Laboratory at Colorado State University. Since January 2006, he has been a postdoctoral research associate with The Program in Applied and Computational Mathematics at Princeton University. His research interests are in statistical signal processing and its applications to distributed sensing, networking, and biomedical imaging.
Friday, February 15th at 2 p.m.
Bossone 302
|
