Maximum a posteriori adjustment of adaptive transversal filters in active noise control

This paper develops a novel approach to adaptive active noise control based on the theory of Bayesian estimation. Control system parameters are considered as statistical variables and a formulation for the joint probability density function of them is derived. An optimal solution for the system parameters is then calculated through maximizing the density function. An efficient adaptive algorithm for iterative calculation of the optimal parameters is proposed. It is shown that the well-known FxLMS algorithm can be derived as a special case of the proposed algorithm, where the noise to be canceled is a white Gaussian process. Simulation results verify the preference of the proposed system to the traditional active noise control systems in terms of steady-state performance and convergence rate. It is also shown that the preference of the proposed system is much more evident when the noise to be canceled is not white. Finally, a successful implementation of the proposed system in an experimental acoustic duct is reported.