ACCELERATING THE CONVERGENCE OF THE FILTERED-X LMS ALGORITHM THROUGH TRANSFORM-DOMAIN OPTIMIZATION

This paper presents an investigation into the use of a transform-domain optimisation, to accelerate the convergence of the filtered-x least mean squares (LMS) algorithm. It is illustrated by results from a real-time active noise control system, cancelling the noise propagating in a one-dimensional duct. On this real-time test platform, we compare the convergence behaviour of the transform-domain filtered-x LMS, to the convergence behaviour of the standard filtered-x LMS. It is shown that the near optimal convergence speed of the transform-domain filtered-x LMS algorithm, its structural simplicity, and its low computational complexity make it a good alternative to other fast algorithms such as the recursive least squares algorithm, the fast Kalman algorithm, or other approaches based on lattice filters. (C) 1995 Academic Press Limited