Final Project Report for University of Notre Dame, Department of Electrical and Computer Engineering, Contract N00014-87-K-0284

Abstract : Recent advances in integrated circuits and signal processing technology have prompted the need of algorithms with higher degrees of modularity and pipeline-ability, resulting in higher degrees of concurrency and machine perception. It is generally believed in the research community that higher degrees of concurrency and machine perception will bring forth many breakthroughs in many signal processing areas, particularly in adaptive systems, speech and image processing and recognition. The objective of this research project is to develop an adaptive signal processing architecture with important features such as modularity, pipeline-ability, and intelligent use of information. The ground work upon which this research project rests is a recursive parameter estimation algorithm, i.e., the so-called OBE algorithm, which features a discerning update strategy. This discerning update is in sharp contrast to the continual update used by most existing algorithms. The estimation algorithm has been developed with a set-theoretic framework. In particular, starting with the assumption that the underlying noise (of the system being studied) is bounded in magnitude, a recursive least-squares type of estimation algorithm was obtained with a discerning update strategy. An important outcome of such discerning updates is that the resulting algorithm can be implemented with two modules: An information processor followed by an updating processor.