A modified frequency domain adaptive filter for active noise control

Frequency domain adaptive filter has potential benefits of fast convergence and low computational load, which make it attractive in the application of active noise control. However, it has been noticed that the commonly used normalized frequency domain block LMS algorithm faces the problem of deterioration of steady-state behavior when the whole control system is noncausal or the adaptive filter is of deficient length. In this paper, an efficient modification of the frequency domain block LMS algorithm is described, and it can be theoretically proven that the proposed algorithm can unconditionally converge to the optimal Wiener solution. Moreover, the convergence behavior is also discussed. The experiment in a short duct active noise control system clearly demonstrate the superiority of the proposed algorithm.