NONLINEAR SIGNAL PROCESSING USING DYNAMICAL SYSTEMS

THE PROPOSED NONLINEAR DYNAMICS METHODOLOGY PROMISES A FOUNDATION FOR A REVOLUTIONARY MICRO-ELECTRONIC TECHNOLOGY THAT COULD PERFORM HIGHLY PARALLEL REAL TIME SIGNAL PROCESSING TASKS SIMILAR TO FOURIER SPECTROGRAPHY WITHOUT A MICROPROCESSOR. IN ADDITION TO FREQUENCY COMPONENT EXTRACTION, THE PROPOSED NONLINEAR SENSORS INTEGRATE AND ENCODE SIGNAL TIME-HISTORY BY UTILIZING JUMP PHENOMENA. THE DEVELOPMENT OF A NEW GENERATION OF SENSOR AND SIGNAL PROCESSING SYSTEMS THAT TRANSDUCE AND ENCODE TEMPORAL INFORMATION IN COMPLEX SIGNALS WOULD BE INVALUABLE, MOST PARTICULARLY IN RECOGNITION TASKS WHERE SIGNAL COMPONENTS HAVE CHARACTERISTIC TIME-EVOLUATIONS AND VARIANCES. THIS PROPOSAL INCLUDES PRELIMINARY EXPERIEMNTAL RESULTS WHICH DEMONSTRATE THE UTILITY OF THIS NEW COMPUTATIONAL FRAMEWORK AS APPLIED TO TWO DIMENSIONAL (ACOUSTIC: SIGNAL VS. TIME) AND THREE DIMENSIONAL (128X128X256-GREY-SCALE IMAGES) SIGNAL PROCESSING TASKS. OVERALL, THE PROPOSED WORK WOULD INTRODUCE A FOUNDATION FOR A NEW SENSOR AND SIGNAL PROCESSING TECHNOLOGY OF GREAT VALUE THAT CLOSELY PARALLELS AND AUGMENTS ONGOING NONLINEAR SCIENCE AND MICROELECTRONICS RESEARCH AT CALTECH AND THE JET PROPULSION LABORATORY.