Birds World: a coupled artificial neural network and expert system (CANNES) architecture

Summary form only given, as follows. The backpropagation neural networks and an expert system are coupled in the system ´Birds World´ for pattern classification. This system fully utilizes both aural and visual data from birds to perform classification tasks at two levels of recognition. The neural network performs an initial classification that is validated by the expert system. Once accepted, the expert system applies its knowledge base for further discrimination within the class. This system optimizes classification performance by incorporating reasoning capabilities for validation as well as for providing further decision support to the neural network. The benefit to the expert system is reflected in a low-overhead base of simple, noncomputational rules.<>