Robust reproduction of sound zones with local sound orientation

Abstract

Pressure matching (PM) and planarity control (PC) methods can be used to reproduce local sound with a certain orientation at the listening zone, while suppressing the sound energy at the quiet zone. In this letter, regularized PM and PC, incorporating coarse error estimation, are introduced to increase the robustness in non-ideal reproduction scenarios. Facilitated by this, the interaction between regularization, robustness, (tuned) personal audio optimization, and local directional performance is explored. Simulations show that under certain conditions, PC and weighted PM achieve comparable performance, while PC is more robust to a poorly selected regularization parameter.

Robust Personal Audio Geometry Optimization in the SVD-Based Modal Domain

“…The acoustic-modelling based regularized personal audio solutions are formulated in [16], [19]. Considering the additive error existing in reproduction systems and the errors in different transfer functions with independent and identically distribution, the additive error is defined as…”

“…where k Q,Q represents the solution solely controls the quiet zone that is expressed with the quiet zone modes. Expressed with the modes of the corresponding sound zone, the solutions (15) and (16) to control a solo zone can be derived as…”

“…ACC maximizes the acoustic energy contrast between the pre-defined listening and quiet zones, PM minimizes error between the reproduced and the desired sound pressures at the controlled points, and PC maximizes correlation between the reproduced sound field and the preferred directional components of the local sound field. The performance of ACC, PM and PC has been compared and reviewed in [11], [16].…”

“…The literature [10], [18], [22] reviewed in detail the development of robustness control in personal audio. Our previous work [16], [22] showed that an acoustic-modelling based regularization can increase a personal audio system's robustness with coarse prior-information, and can be further employed to system implementation with less in-situ measurement of transfer functions.…”

“…Methods using objective [18] or perceptual metrics [28] have been studied for choosing loudspeaker locations with the array effort based regularization. From our previous work [22] and [16], the array effort based regularization has difficulty to determine the threshold and sometimes cannot provide proper regularization.…”