OPTICAL COMPUTING AND OPTICAL SIGANL PROCESSING

THE REDUCED LEVEL OF INTERFERENCE IN OPTICAL COMPUTERS AND THE LOW THERMAL MASS OF THEIR MATERIALS MAKE OPTICAL COMPUTING SIGNAL PROCESSORS AN ATTRACTIVE ALTERNATIVE FOR STRATEGIC APPLICATIONS. THE NEURAL-NETWORK APPLIED TO SIGNAL AND IMAGE PROCESSING HAS BEEN DEMONSTRATED TO EXHIBIT COMPARABLE PERFORMANCE TO THAT ACHIEVABLE USING MORE CONVENTIONAL APPROACHES. OPERATIONAL AND PERFORMANCE REQUIREMENTS ARE BEING DEFINED FOR THE NEURAL NETWORK SIGNAL PROCESSOR (NNSP). THIS FEASIBILITY IS BEING INVESTIGATED OF THE NNSP CONCEPT IN BOTH ITS ELECTRONIC AND OPTICAL IMPLEMENTATIONS. THE NEURAL NETWORK APPROACH IS BEING COMPARATIVELY ANALYZED WITH STANDARD METHODS INTERMS OF PERFORMANCE LEVEL, COMPUTATIONAL COMPLEXITY AND TECHNICAL RISK. THE APPROPRIATE NNSP MODEL IS BEING DEFINED FOR A PROOF-OFPRINCIPLE DEMONSTRATION. THE IMPLEMENTATION OF THIS SIGNAL PROCESSINGMODEL INTO OPTICS IS BEING EXPLORED INCLUDING THE DEFINITION OF THE INPUT-OUTPUT INTERFACES BETWEEN THE OPTICAL NNSP AND THE LONG WAVELENGTH INFRARED SENSOR FOCAL PLANE AND PROCESSOR. DESIGN TRADE STUDIED FOR THIS DEVICE ARE BEING PERFORMED AND A DETAILED DESCRIPTION PROVIDED OF A METHODOLOGY LEADING TO THE FABRICATION OF A PROTOTYPE OPTICAL NNSP. THE POTENTIAL APPLICATIONS FOR OPTICAL COMPUTING AND FOR NEURAL-NETWORKS INCLUDE IMAGE UNDERSTANDING SYSTEMS;COGNITIVE SYSTEMS FOR DECISION SUPPORT APPLICATIONS; AND PATTERN RECOGNITION SYSTEMS IN STABLE TRAINING ENVIRONMENTS.