A STABILIZED FAST TRANSVERSAL FILTERS ALGORITHM FOR RECURSIVE LEAST-SQUARES ADAPTIVE FILTERING

In this paper, the numerical stability of the well known gain-normaliz ed fast transversal filters (GFTF) algorithm is analytically investiga ted. Starting from an analysis of the error propagation, suitable roun d-off errors difference equations (REDE) are derived. The stabilizatio n of the GFTF algorithm can be achieved by stabilizing the REDE. The r esulting algorithm is numerically stable, having only 7N + 27 multipli cations and divisions per recursion (MADPR). The numerical stability o f the proposed algorithm is demonstrated by the simulation results.