COMPUTER AIDED DESIGN (CAD) DIGITAL FILTERING/SMOOTHING SCHEMES

RECENTLY CANONICAL VARIATE ANALYSIS (CVA) THEORY HAS BEEN SUCCESSFULLY APPLIED TO ADAPTIVE IDENTIFICATION OF MULTI-INPUT MULTI-OUTPUT (MIMO) SYSTEMS IN OPEN - AND CLOSED-LOOP CONFIGURATION. THE METHOD AUTOMATICALLY SELECTS AN OPTIMAL MODEL ORDER BASED ON AN INFORMATION THEORETIC CRITERION KNOWN AS AKAIKE INFORMATION CRITERION (AIC). UNDER THE SCOPE OF THIS PROPOSAL, A NEW FILTERING AND SMOOTHING SCHEME BASED ON CVA-AIC WILL BE DEVELOPED THAT WILL BE USED TO PROCESS DATA WITH ABRUPT CHANGES. THE PHASE I EFFORT OF A THREE PHASE RESEARCH PROGRAM CONSISTS OF THREE MAJOR TASKS: (I) TO DEVELOP A CVA-AIC BASE FILTERING/SMOOTHING SCHEME FOR HANDLING DATA WITHOUT ANY ABRUPT CHANGE, (II) TO DEVELOP A CVA-AIC BASED TECHNIQUE FOR DETECTING AND ESTIMATING THE TIME OF OCCURRENCE OF ABRUPT CHANGE AND (III) INTEGRATING THESE SCHEMES INTO A COMPLETE DESIGN PACKAGE. CVA-AIC TECHNIQUE WILL ALSO BE USED FOR OPTIMAL ORDER SELECTION OF THE POLYNOMIAL IN THE "CLASSICAL" LEAST SQUARE MOVING ARC POLYNOMIAL SMOOTHING TECHNIQUES. PHASE I WILL LAY THE FOUNDATION FOR PHASE II DURING WHICH A CVA-AIC BASED CAD SOFTWARE PACKAGE FOR DESIGNING FILTERING/SMOOTHING SCHEMES WILL BE DEVELOPED.