A diffusion filtered-x affine projection algorithm for distributed active noise control

Abstract

The well-known Filtered-x Affine Projection (FxAP) algorithm is usually deemed a better choice than the conventional Filtered-x Least-Mean-Square (FxLMS) algorithm when faster convergence speed is desired in active noise control applications. However, the improvement of its convergence performance is obtained at the expense of increased computational complexity. It is usually unrealistic to regulate multichannel ANC systems using the centralized AP algorithm. Recently, distributed diffusion adaptation schemes over acoustic sensor and actuator networks have been introduced to multichannel ANC systems, such as the Diffusion FxLMS algorithms. Distributed processing based on diffusion cooperation strategy allocates the computation load to the network nodes in a scalable manner. This paper proposes a distributed Diffusion Filtered-x AP (DFxAP) algorithm for multichannel ANC systems. Simulation results and computational complexity analysis show that the DFxAP algorithm outperforms the Diffusion FxLMS algorithm in convergence performance and has lower computational complexity compared to the centralized AP algorithm.