摘要
Global control of noise in a three-dimensional space is generally difficult, given a limited number of discrete sensors. This paper presents a feedforward multichannel active noise control (ANC) approach to overcome this difficulty. In light of the model-matching principle, an underdetermined multichannel inverse filtering (UMIF) system is formulated. Infinite number of solutions exist to yield zero residual noise at the error microphone positions. By incorporating the UMIF system as a design constraint, a multichannel feedforward controller is formulated using the linearly constrained minimum variance (LCMV) approach in which a cost function is introduced to minimize the residual noise energy at a large number of preselected fictitious control points. To implement the ANC system in reverberant environments, a novel sensor interpolation technique based on plane wave decomposition is developed to find the frequency responses between secondary loudspeakers and the fictitious control points. Only a limited number of room frequency response measurements are required in the use of these techniques. Simulations and experiments are undertaken using a uniform linear loudspeaker array to validate the proposed ANC system. The results have demonstrated that the proposed approach yielded substantial noise reduction in a widened and connected control area.