Electro-optical Fuzzy Pattern Recognition And Signal Processing

THE PROPOSED INNOVATION, A SINGLE STAGE, FULL DUPLEX, ELECTRO- OPTICAL SYSTEM, WILL IMPROVE AND SPEED UP THE CLASSIFICATION AND SIGNAL PROCESSING OF DIFFERENT TYPES OF PATTERNS USED BY THE SDI ORGANIZATION. WHEN OUR ELECTRO-OPTICAL SYSTEM USED IN THE FORWARD DIRECTION, IT PERFORMS FUZZY PATTERN RECOGNITION BY TRAINING ALL THE CLASSES IN PARALLEL. SIMULTANEOUSLY THE SAME SYSTEM CAN BE USED IN THE REVERSE DIRECTION TO DO BANDPASS AND/OR IN-BAND FILTERING BY PROCESSING A NUMBER OF NOISY INPUT FRAMES IN PARALLEL. IN THE PAST DECADE ARTIFICIAL NEURAL NETWORKS HAVE BEEN USED TO CLASSIFY PATTERNS IN MANY FIELDS. HOWEVER, PROBLEMS HAVE BEEN ENCOUNTERED IN INTELLIGENT SYSTEMS, E.G., MACHINE VISION PLATFORMS THAT ATTEMPT TO INTERPRET SCENES WITH VARIOUS KINDS OF SENSORS AND CONTEXTUAL INFORMATION, BY THE INCORRECT CLASSIFICATION OF TEST PATTERNS. IMAGINE AN SDI PATTERN INDICATING THREE DIFFERENT SIGNATURES. THE FIRST ONE, HOWEVER, IS MORE PRONOUNCED THAN THE OTHER TWO. IN SUCH A CASE THE NEURAL NETWORK WILL IDENTIFY THE FIRST SIGNATURE BECAUSE IT ACHIEVED THE MINIMUM ERROR FOR THIS PARTICULAR CLASS AND IGNORE THE OTHER TWO. AS A RESULT, THE EXPERT SYSTEM OF THE INTELLIGENT SYSTEM WILL EITHER FAIL TO IDENTIFY OR WRONGLY INTERPRET THE SITUATION AND MAKE THE WRONG DECISION. THE USE OF FUZZY RATHER CRISP CLASSIFICATION CAN AVOID THESE PROBLEMS. 518 WE PROPOSE A NOVEL METHOD FOR PROVIDING OPTICAL MODULATOR PADS ON SILICON VLST CIRCUITS. OUR TECHNIQUE SHOULD ENABLE MODULATOR PAD ARRAYS OF NEARLY ARBITRARY CONFIGURATION, WITH MINIMUM PAD GEOMETRIES OF A FEW MICRONS SQUARE. MODULATOR RESPONSE TIMES LESS THAN 1 ns SHOULD BE ACHIEVED. THE TECHNIQUE REQUIRES ONLY A FEW PROCESS STEPS AND WILL WORK ON CIRCUITS MADE IN ANY STANDARD VLSI PROCESS.