Practical techniques for rapid and reliable real-time adaptive filtering

Simplicity, flexibility, and reliability are three important aspects of practical adaptive filtering systems. In this work, two techniques are investigated which address these issues. First, a new class of data-reusing LMS algorithms is explored. These algorithms are seen, through extensive simulation examples, to have superior convergence rate and Mean-Squared Error performance over the Data-Reusing LMS algorithm at the same computational cost. A geometric framework which aids in the presentation of the new class of algorithms is developed. This framework also allows a more complete understanding of three existing LMS-type algorithms, as well as allows the proof of several important convergence rate properties which relate the three algorithms.Second, a novel fault tolerance mechanism called Adaptive Fault Tolerance (AFT) is introduced. This fault tolerance approach is then applied to Finite Impulse Response (FIR) adaptive filters. The Fault Tolerant Adaptive Filters (FTAFs) which result from using AFT are analyzed with respect to their convergence rate, computational complexity, and hardware overhead. The goal of this investigation is to develop a practical and useful FTAF, which can tolerate numerous coefficient failures regardless of the input noise statistics. Adaptive Fault Tolerance provides protection against many coefficient faults with very low hardware overhead.U of I OnlyETDs are only available to UIUC Users without author permissio