An Exact Relaxed Fast Affine Projection Algorithm for Multichannel Active Noise Control

In the field of adaptive filtering, affine projection algorithms and their fast implementations are known to produce a good tradeoff between complexity and convergence performance. For multichannel active noise control, some fast affine projection algorithms have consequently been introduced in recent years. However, these previously published algorithms are either based on non-relaxed versions of the fast affine projection algorithm, or on sub-optimal approximations of the algorithm. Non-relaxed versions of the fast affine projection algorithm produce a sub-optimal performance when non-unity step size values are used, which are sometimes desirable in practice. Moreover, not being exact, the previously published algorithms are rather sensitive to the regularization factors being used, making them more difficult to tune and less robust in practice. In this paper, an exact relaxed fast affine projection algorithm for multichannel active noise control is introduced. Simulation results validate the improved convergence of the relaxed algorithm for non-unity step size values, and the good convergence for a much wider range of regularization factors.