An intermittent RLS algorithm and its application to adaptive active wideband noise control

The most commonly used algorithm for an adaptive active acoustic noise controller is the filtered-x least mean square (FLMS) algorithm which has been applied successfully to active pure tone or narrowband noise control. However, the FLMS algorithm fails in active wideband noise control because of slow convergence rate and sensitivity to dispersion of eigenvalues of the noise autocorrelation matrix. Based on the principle of recursive least squares, an intermittent RLS (IRLS) algorithm is presented, which updates recursively weight coefficients of an adaptive filter at intervals. The intermittent period equals the time delay when the sound wave propagates from the secondary source to the error sensor. Analytic and simulation results show that the IRLS algorithm has excellent convergence behavior, stability and high attenuation level (AL) in comparison with the FLMS algorithm, hence this algorithm is particularly suitable for active wideband noise control. Finally, the feasibility of hardware implementation of the IRLS algorithm is discussed.<
>