The problem addressed is that of extracting information about events associated with a physical system by processing a large number of data streams flowing from sensors and instrumentation. It is assumed that deceptive data portions are imbedded, with no a-priori models describing system behavior. The objective is to formalize, articulate, and demonstrate a signal-processing framework (object oriented signal processing) which is emerging from experience with problems of this nature in commercial applications. In this setting instrumentation injects raw material into a material-processing framework, with a rigorous mathematical platform for developing condensation and classification processes. Condensation processes extract information from raw-material streams and inject it into highly compressed (1000:1) streams of synthetic material feeding classification processes. High performance gains result from applying intensive processing directly to compressed material. Classification processes produce groups of similar material characterized by distributions of material populations. There appears to be a close relationship between these distributions and membership functions as used in fuzzy control.
