Distributed and quantized weak signal detection
Abstract
This dissertation begins with a tutorial survey which discusses many of the important topics associated with distributed detection. An extensive bibliography is provided which is organized by subject headings. Next, locally optimum distributed detection theory is investigated in some detail for the case of observations which are possibly dependent from sensor to sensor. The results focus on the case of two sensors but extensions to more than two sensors are provided. A key example for detection of dependent random signals in additive and possibly non-Gaussian noise shows that locally optimum procedures may process observations differently from sensor to sensor for identical observation models. The asymptotically optimum design of a particular distributed detection system, which takes a single independent observation from each sensor, is shown to be mathematically equivalent to the asymptotically optimum design of a signal detection scheme based on quantized observations taken at different time instants. Design examples for this distributed detection problem focus on solutions under the constraint of identical sensor processors. One application of our locally optimum distributed detection results for possibly dependent observations is given in which a cell-averaging constant false alarm rate detection scheme is applied to a problem with two sensors which receive exponential observations.
Subject Area
Electrical engineering|Systems design
Recommended Citation
Blum, Rick S, "Distributed and quantized weak signal detection" (1991). Dissertations available from ProQuest. AAI9211908.
https://repository.upenn.edu/dissertations/AAI9211908