A Reduced Complexity Random Fourier Filter Based Nonlinear Multichannel Narrowband Active Noise Control System

Linear-in-the-parameters nonlinear controllers are widely used for active noise cancellation (ANC) in the presence of nonlinearities in the system. However, the performance of such controllers deteriorates significantly in highly nonlinear systems and also in the presence of nonlinear harmonic distortion. To overcome these issues, a kernel filtered-x least mean square (KFxLMS) algorithm based narrowband ANC system has been recently reported in the literature. We observe that the large memory size and high computational complexity of the KFxLMS algorithm hinder its implementation in real-time ANC systems. To alleviate these issues, a single channel random Fourier filter based FxLMS (RFFxLMS) algorithm is proposed in this brief. The above formulation is then extended to the multichannel scenario, and an RFF based multichannel narrowband ANC system is developed to achieve noise reduction over a broader area. The computational load of the proposed multichannel approach is further reduced by using a partial updating method. The simulation study shows that the overall performance of the proposed schemes is significantly better than the existing nonlinear methods.

Automated summary

A narrowband ANC system based on kernel filtering and random Fourier filters is proposed to address the performance deterioration of nonlinear controllers. The approach is extended to multichannel and employs partial updates to reduce computational load, outperforming existing nonlinear methods in simulation studies.