MULTILEVEL SPLIT-PATH ADAPTIVE FILTERING AND ITS UNIFICATION WITH DISCRETE WALSH TRANSFORM ADAPTATION

It has been shown that the convergence speed of an adaptive transversa l filter can be made faster if the filter is split into a pair of symm etric/antisymmetric linear phase subfilters. In this brief, a generali zed multilevel splitting approach is proposed which enables the realiz ation of a modular split-path adaptive filter. It is demonstrated that an LMS FIR filter of length 2L, where L is an integer, can be decompo sed into a set of parallel single-parameter subfilters by a novel L-st ep splitting procedure. The convergence behavior of the new filter con figuration is improved noticeably and is confirmed experimentally. A u nified perspective of the multilevel split-path adaptive filtering met hod with the discrete Walsh transform adaptation is also established.