2018 年 84 巻 859 号 p. 17-00446
Because of using interference of sound wave, it is difficult to control in whole three-dimensional acoustic field by Active Noise Control (ANC). Instead, around-head-control is investigated in this paper. In this system, two error microphones are set near ears, and noise is reduced around evaluation points locally. This system requires fast adaptation speed of controller following movement of head, and a problem still remains that the control effect during movement becomes worse. Against this problem, we propose Modified Reference signal Method (MRM). This method improves adaptation speed during movement, by modification of filtering of input signal in Filtered-x. The secondary path model that controller requires is interpolated by the method Ohno et al. proposed. There is modeling error of secondary path caused by interpolation and measuring error of it, and this modeling error causes divergence of controller. In this paper, suppression method of divergence is also investigated. Suppression of divergence is executed by addition of penalty term to evaluation value, and it is equivalent to ridge regression mathematically. Proposed method requires a large number of matrix calculations, therefore reducing method of computational complexity according to acoustic characteristics is also investigated in this paper. The validity of the proposed method is shown by numerical simulation and experiment.
