A modified filtered-x LMS algorithm for active control of periodic noise with on-line cancellation path modelling

The filtered-x LMS algorithm (FXLMS) has been successfully applied to the active control of periodic and random noise and vibration. This paper presents a modified algorithm for active control of periodic noise based on the FXLMS algorithm which uses random noise for on-line cancellation path transfer function (CPTF) estimation. In the proposed algorithm, another two short adaptive filters are introduced. One is an adaptive noise cancellation filter; which is used to improve the convergence speed of the CPTF modelling filter in the presence of very large amplitude primary noise hy cancelling the component of the error signal that is correlated with the primary noise. The other is an adaptive estimator which is used to re-estimate the obtained CPTF (long FIR filter estimated by random noise) with a short FIR filter by using the periodic reference signal as the input. The traditional FXLMS algorithm is then used with the shortened FIR filter to filter the reference signal, thus providing significant processing flexibility in practical situations where the primary path transfer function changes much faster than the CPTF. Simulation results demonstrate the effectiveness of the proposed algorithm.