In beamforming-based sound field acquisition and reproduction systems, directional sound acquisition is first performed using beamforming. Then, the obtained signals are used to drive virtual sound sources corresponding to each beam direction, thereby reproducing the sound field. However, due to the presence of sidelobes in typical beam patterns, the acquired signals are not strictly directionally isolated, which degrades spatial reproduction accuracy. Therefore, conventional methods address this issue by applying weighted linear combinations of beam signals so that the original sound field can be reproduced. We propose a method for designing beamforming filters by setting a window function as the desired beam output. This approach effectively suppresses sidelobes, as confirmed by our evaluation. To evaluate the performance of this window-targeted beamforming (WTBF) method in sound field reproduction, we conducted simulations to generate binaural signals representing a scenario with two sound sources. Subsequently, a listening experiment was performed to compare the perceived spatial impression with that of the original sound field. The results indicate that, compared to conventional methods, our proposed approach can present a spatial impression close to the original without applying additional weighting, thereby demonstrating the superiority of the WTBF method.
Watanabe et al. (Wed,) studied this question.