Symmetric adaptive predictive structure for tracking channel non-stationarities

Abstract

The basic idea of this paper is related to the fact that the steady state property in a non-stationary system context is strongly related to the input correlation characteristics. When the LMS algorithm is used to identify a system with variations modeled by a random walk, the performance is degradated as the input correlation increases. The classical identification scheme can be improved by the use of a prewhitening adaptive filter. A theoretical analysis of an adaptive predictive identification scheme is presented. This study illustrates the contribution of predictive structures for tracking system non-stationarities.