Reduced-complexity decentralized detection of spatially non-constant phenomena

In this chapter, we study sensor networks with decentralized detection of a non-constant phenomenon, whose status might change independently from sensor to sensor. In particular, we consider binary phenomena characterized by a fixed number of status changes (from state “0” to state “1”) across the sensors. This is realistic for sensor networking scenarios where abrupt spatial variations of the phenomenon under observation need to be estimated, e.g., an abrupt temperature increase, as could be the case in the presence of a fire in a specific zone of the monitored surface. In such scenarios, we derive the minimum mean square error (MMSE) fusion algorithm at the access point (AP). The improvement brought by the use of quantization at the sensors is investigated. Finally, we derive simplified (sub-optimum) fusion algorithms at the AP, with a computational complexity lower than that of schemes with MMSE fusion at the AP