Title :
Analysis of the filtered-X LMS algorithm
Author :
Bjarnason, Elias
Author_Institution :
Tech. Hochschule Darmstadt, Germany
Abstract :
The presence of a transfer function in the error path, as in the case of active noise control requires a filtered update of the LMS (least mean square) algorithm. To ensure convergence of the algorithm, the input to the error correlator has to be filtered by a copy of the error path transfer function. The author presents an analysis of the filtered-X LMS algorithm in the presence of spherically invariant input process. Some theoretical results and simulated behavior are shown to agree very well.<>
Keywords :
convergence of numerical methods; correlators; filtering and prediction theory; least squares approximations; transfer functions; active noise control; convergence; error correlator; error path transfer function; filtered-X LMS algorithm; least mean square; spherically invariant input process;
Conference_Titel :
Acoustics, Speech, and Signal Processing, 1993. ICASSP-93., 1993 IEEE International Conference on
Conference_Location :
Minneapolis, MN, USA
Print_ISBN :
0-7803-7402-9
DOI :
10.1109/ICASSP.1993.319547
