A Family of Equation-Error Based IIR Adaptive Algorithms

P.S.R. DINIZ; J. E. COUSSEAU

Lafayette, LA

Congreso; 37th Midwest Symposium on Circuits and Systems; 1994

IEEE Circuits and Systems Society

The algorithm for adaptive IIR filtering based on the Least Meun Squares Equation Error (LMSEE) has desirable properties in terms of stability and unimodality of the error surface. The basic problem related to this method is the parameter bias induced by the measurement noise [1]-[a]. Alternative solutions proposed to solve this problem are the InstrumentalVariable methods [3]. In these methods the stability can not be guaranteed under the same general conditions as for the LMSEE method. In this paper is presented a family of asymptotically stable algorithms, based on a modification of the basic LMSEE updatingequations, that within sufficiently general conditions leads to consistent parameter estimates. After introducing the family of algorithms, we analyse its basic properties such as: stationary points, stability and convergence of the estimates obtained for each memberof the farnily.