Optimality for data fusion of detectors using Voronoi cells

The problem of optimal signal classification with distributed sensors and restricted communication channels is considered. The classifier is to decide if a signal is in one of several possible classes after being observed through a suite of sensors, each sensor giving rise to a soft classification. The local classifiers are assumed to be based on a Voronoi decomposition of the associated sensor output space. Conditions of optimality for both the classifier and the sensors are given.