Improved LMS adaptive algorithm with varying stepsize

摘要

In the field of system identification, the coefficients of a digital filter, which are used to illustrate the transfer function of a system, are often fixed by known input and output signal sequences. The least mean square (LMS) adaptive algorithm isoften used for its small computational requirement and convergence performance. Generally, the input or the output signal sequence often contains directive current (DC) components. What is the effect of the DC component upon the adaptive algorithm? Thisis just the main objective of this note. Here we propose an improved LMS adaptive algorithm with varying stepsize, which is based on the DC components analyzing.