Chou C., Li H., Chang C.

In this paper we propose a new method to attenuate predictable noise components inside the desired signal by using adaptive feedback active noise control with on-line secondary-path modeling. The existing adaptive feedback active noise control methods are using off-line secondary-path modeling which are inefficient for time-varying systems. The proposed method is developed based on the Filtered-x Least Mean Square (FxLMS) algorithm, and the secondary path is on-line estimated by adding additional random noise into the ANC system. The error signal is feedback mixing with the signals from the controller and the secondary path estimator to generate the reference signal. The incoherent disturbances in each residual error of the cross-updated least mean square adaptive filters are mutually eliminated when attenuation of the desired signal and estimation of the secondary path are operating simultaneously. Furthermore, the synthesized desired signal converges to the desired signal during operation as well. Performances of narrowband and broadband active noise control (ANC) in some simulating system are conducted and the results show that the predictable noise components are attenuated.