Document Type
Thesis
Date of Award
5-31-1989
Degree Name
Master of Science in Electrical Engineering - (M.S.)
Department
Electrical Engineering
First Advisor
Yeheskel Bar-Ness
Second Advisor
Chung H. Lu
Abstract
Adaptive systems using the LMS algorithm or its variants usually require bilinear correlation multipliers with high dynamic range and large bandwidth. The class of LMS-type algorithms utilizing zero-memory limiters (on one or both correlation multiplier inputs) to reduce dynamic range are examined. The correlator limiters may be included to simplify the hardware implementation or may arise due to limited dynamic range of the loop components. The application of a theorem due to Price provides a technique for simplifying the analysis, which otherwise might be extremely difficult. The analysis provides closed-form expressions for the mean weight convergence behavior, permitting examination of the convergence bounds and performance for this class of algorithms.
Recommended Citation
Van Saders, John George, "Nonlinear variants of the least mean square (LMS) algorithm" (1989). Theses. 2939.
https://digitalcommons.njit.edu/theses/2939
