|Table of Contents|

The Study of Steady State Square Error in the LMS Algorithm(PDF)


Research Field:
Publishing date:


The Study of Steady State Square Error in the LMS Algorithm
Wang Yunhong Liu Guosui Wang Yiding
School of Electronic Engineering and Optoelectronic Technology, NUST, Nanjing 210094
adapt ive f iltering w eig ht ( mathematics) errors
T his paper discusses the relationship amo ng the co rrelation of input sig nal , w eig ht s number and steady state squar e err or of the LMS alg orithm in detail. It has been pr oved that the increasement o f w eig ht s number can not ensur e the reductio n of steady state square erro r. For pract ical input signal, there is an optimum w eig hts number w hich makes the steady state er ror the least value. This error may be increased w hen the w eig ht s number is increased. A direct current signal and a sine signal are used as the input s o f the adaptive filter to do the exper iment by computer simulat ion. T he experiment result s are in go od ag reement w ith theoret ical ones, T he theor y in this paper pr ovides the guidance of adaptive filter design in pract ice.


1 Hay kin S. Adaptive filter thero y . Second Ed Englewo od Cliffs, N J, Prentice-Hall, 1991
2 Winrow B, Dstear ns S. Adaptiv e sig nal pr ocessing . Eng lew oo d cliffs, N J: Pr enice-Hall,I nc, 1985
3 Gar dner W A. Nonstationar y lea rning char acteristics of the LMS alg or it hm s. IEEE Tr ans,1987, CAS34( 10) : 1199~1207
4 Widrow B, Mccoo l J M. Stationar y and no nstat ionar y lear ning char acter istics o f the LM S adaptiv e filter . Pr oc IEEE, 1976, 64( 8) : 1151~1162


Last Update: 2013-04-11