Sensor Networking for Detection: From Distributed Detection to Energy Savings, MIMO Radar and Image Fusion

Sensor networking has become more popular in recent years as a powerful approach in both the military and commercial sectors to monitor various critical events involving, for example, weather, physical structures, animals and biology. Many of these monitoring operations involve hypothesis testing, often called signal detection, (or classification) to determine if various events have occurred so that required actions can be initiated. Here we review some interesting theoretical aspects of sensor networking for signal detection problems when the topic is considered from some basic, but very different, formulations. First, we consider a distributed detection problem, where each distributed sensor will make a hard (or multi-bit) decision about a hypothesis testing problem and these decisions will be combined to make an overall decision. If the sensor decisions are statistically independent when conditioned on the hypothesis, the problem is shown to be relatively simple when compared to statistically dependent cases. We interpret some statistically dependent solutions. We next consider the problem of saving energy from reducing wireless transmissions by using censoring or ordering. Unlike censoring, ordering can provide energy savings while still achieving the optimum unconstrained hypothesis testing performance. If the sensors in the network are widely spread transmit and receive antennas, they can be used to form an active or passive MIMO radar. Some basic ideas of MIMO radar with widely spaced antennas are discussed. Finally, if the sensor nodes are cameras with various modalities, image fusion can be accomplished over the network. Some basic motivations and aspects of image fusion are discussed.