Title :
Performance analysis of time domain block LMS algorithm
Author :
Hsieh, Luke S L ; Wood, Sally L.
Author_Institution :
National Semiconductor Corp., Santa Clara, CA, USA
Abstract :
The performance advantages of the time-domain block LMS (least mean squares) algorithm (BLMS) relative to the single update LMS algorithm are studied. An exact relationship of weight update between BLMS and LMS is presented. Based on this relationship an equation of equivalent mu /sup B/ for mu is derived. The performance of BLMS is analyzed in terms of convergence rate, misadjustment, signal to noise ratio, and stability bound, and compared with that of LMS in terms of the equivalent mu /sup B/. Simulation results support the analysis. Modifications which could improve performance are explored. This analysis of BLMS and LMS provides a basis for modified block algorithms.<>
Keywords :
adaptive filters; convergence of numerical methods; filtering and prediction theory; least squares approximations; stability; time-domain analysis; convergence rate; least mean squares; misadjustment; performance advantages; signal to noise ratio; stability bound; time domain block LMS algorithm; weight update;
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.1993.319553
