MULTISENSOR INFORMATION PROCESSING USING A MULTIPLE NEURAL NETWORK LEARNING SYSTEM

THE OBJECTIVE IS TO DEMONSTRATE THE FEASIBILITY OF USING NESTOR'S NEURAL NETWORK LEARNING TECHNOLOGY FOR PATTERN RECOGNITION IN A MULTISENSOR AVIONICS SYSTEM ENVIRONMENT. PATTERN MATCHING VIA PARALLEL SEARCHES IN A MULTISENSOR DATA ENVIRONMENT MUST BE CONTROLLED BY AN APPROPRIATE HIERARCHICAL DECISION LOGIC THROUGH WHICH A SYSTEM CAN LEARN HOW BEST TO USE ITS AVAILABLE DATA FOR ROBUST RECOGNITION. NESTOR HAS PIONEERED A NOVEL SYSTEMS APPROACH TO USING MULTIPLE NEURAL NETWORKS FOR PATTERN RECOGNITION, WHERE EACH COMPONENT NETWORK OF THE SYSTEM CAN BE CONFIGURED TO PROCESS INFORMATION FROM A SINGLE SENSOR OR MULTIPLE SENSORS. A HIGH LEVEL CONTROLLER, OPERATING ON THE SYMBOLIC OUTPUT OF INDIVIDUAL NETWORKS, INTEGRATES THEIR RESPONSES AND DETERMINES FEEDBACK TRAINING SIGNALS. THIS TECHNOLOGY HAS BEEN SUCCESSFULLY APPLIED TO PATTERN RECOGNITION PROBLEMS IN CHARACTER RECOGNITION, VISUAL OBJECT RECOGNITION AND THE SIMULATION OF HUMAN EXPERT DECISION MAKING. WE PROPOSE TO ASSESS THE PERFORMANCE OF THE SYSTEM ON REAL-WORLD MULTISENSOR DATA TO SHOW THE FEASIBILITY OF USING THE TRAINABLE DECISION-MAKING LOGIC OF THE SYSTEM FOR MULTISENSOR PROCESSION IN AN AVIONICS ENVIRONMENT.