SPLIT FILTER STRUCTURES FOR LMS ADAPTIVE FILTERING

In this paper, the performance of a split adaptive filter in either a parallel or a serial form for non-white inputs is investigated. A para llel split structure is constructed by using two linear phase filters connected in parallel while for serial split it is configured as a cas cade of two transversal filters. The adaptation characteristics of the well-known LMS algorithm has been shown to be governed by the eigenva lue spread of the input correlation matrix. By adopting the split stru ctures, we illustrate that the eigenvalue ratios of the associated cov ariance matrices can be reduced thereby giving rise to a faster conver gence speed. The parallel and serial split adaptive filters are examin ed for both joint process estimation and linear prediction. A new type of linear predictor is formed by combining both split methods and its performance for speech analysis is studied. Simulation results are in cluded to validate the superiority of the proposed split filter struct ures in improving the rate of convergence for LMS adaptation.