The gradient adaptive split lattice algorithm

The authors present a new algorithm for adaptive filtering of autoregressive processes, based on a three-term recurrence that replaces the lattice equations of the gradient adaptive lattice (GAL) algorithm of Griffiths (1977) and requires only half as many multiplications per recursion. They also present a single-stage convergence analysis and a numerical example comparing the new algorithm to the GAL. Its performance is similar to the GAL, while requiring less than 2/3 as many multiplications and additions.<
>