The proposed research is directed towards developing a methodology that will ease the modelling and the implementation of multi-channel signal interpretation problems. An object-oriented approach is followed and the system is organized around two separate but collaborating knowledge bases. The first involves the heuristics and the subjective knowledge commonly utilized by expert interpreters, the second uses the more qualitative knowledge utilized by a signal processing specialist. The collaboration between the two knowledge bases are handled by a coupling mechanism. The proposed system permits the modelling of interpretation problems in terms of events and relations between events, and provides a series of operators to express and represent spatio-temporal and contextual information. This research is a first attempt at separating domain- dependent and domain-independent (signal processing) knowledge, thus improving generality, reusability, and facilitating the collaboration of the two types of expertise typically required for the development of automatic signal interpretation systems. Specifically, it advances the understanding of the theory of deep coupled systems where numerical and symbolic methods are combined to improve overall system performance, and addresses the problem of control flow and accumulation of evidence for spatio-temporal reasoning. To demonstrate the generality of the approach, the system is tested on two different applications taken from the medical domain: the interpretation of recordings of the cerebral activity (EEG) and of muscular activity (EMG).
