Title :
A switched regressor algorithm for adaptive IIR filtering
Author :
Burt, Phillip M S
Author_Institution :
LCS-DEE/EPUSP, Sao Paulo, Brazil
Abstract :
In the present work an algorithm for adaptive IIR filtering is proposed, in which the updating of the coefficients can be switched between the Steiglitz-McBride identification method (SMM) and the recursive prediction error method (RPEM), or vice-versa, in order to reach the global minimum of the cost function or its proximity. We also analyse aspects related to the speed of convergence of the SMM and the RPEM, and aspects related to the minimization of the output error in the undermodelled case. This serves to justify the proposed algorithm
Keywords :
IIR filters; adaptive filters; convergence of numerical methods; error analysis; filtering theory; identification; prediction theory; Steiglitz-McBride identification method; adaptive IIR filtering; coefficient updating; convergence speed; cost function; global minimum; output error minimisation; recursive prediction error method; switched regressor algorithm; undermodelled case; Adaptive filters; Convergence; Cost function; Electronic mail; Equations; Error correction; Filtering algorithms; IIR filters; Transfer functions; White noise;
Conference_Titel :
Circuits and Systems, 1995., Proceedings., Proceedings of the 38th Midwest Symposium on
Conference_Location :
Rio de Janeiro
Print_ISBN :
0-7803-2972-4
DOI :
10.1109/MWSCAS.1995.510258
