Combining background noise and artificial masking to achieve privacy in sound zones

Abstract A private sound zone can be created by focusing a spoken message towards a target listener using a loudspeaker array. In practice, however, the reproduced speech cannot be completely contained within the target zone due to practical limits on the directivity of the array. Despite these limitations, the privacy of the message can be maintained if the leaked speech is sufficiently masked by noise. Two possible sources of this masking noise are considered in this article: the ambient noise in the reproduction environment, and an additional masking signal radiated by the loudspeaker array. The present article demonstrates that the process of designing a private audio system is significantly affected by the presence of ambient noise. A key complication is that temporal fluctuations and spatial non-uniformity in the ambient noise can reduce its effectiveness as a masker. These features also make it more difficult to estimate the corresponding reduction in the intelligibility of speech in each listening zone. To mitigate this spatial and temporal variance, it is proposed that systems should be designed to rely only on the masking provided by the diffuse, quasi-stationary background noise component of the environmental noise. It is shown that when systems utilise a combination of the background noise and an additional, artificial masker, a lower level of acoustic contrast is required from the system, compared to the case where the masking is supplied by the background noise exclusively.