On the implementation of fully adaptive interpolated FIR filters

This paper presents a novel strategy for implementing fully adaptive interpolated finite impulse response (FAIFIR) structures using either the least-mean-square (LMS) or the normalized LMS algorithms. The aim of such a strategy is to mitigate numerical stability issues arising from simultaneously adapting the two cascaded filters (sparse filter and interpolator) that compose a FAIFIR structure. In this context, a modification in the structure of the interpolator is proposed with no impact on both the computational complexity and applicability of the FAIFIR structure. As a result, adaptive filters with enhanced numerical properties are obtained. Numerical simulation results are presented attesting the effectiveness of the proposed strategy.