Stochastic modeling of car cabin acoustics with application to personal sound zone systems

Personal sound zone (PSZ) systems aim to provide personalized and individual listening experiences to groups of people who share a common environment. This is achieved through proper control of the sound field, performed via advanced digital signal processing of the audio signals emitted by a set of loudspeakers. Reliable and realistic performance estimation of PSZ systems requires numerous time-consuming measurements of the acoustic responses in the considered environment. In this paper, a measurement campaign of the acoustic responses in a segment-C car cabin was carried out to characterize sound propagation inside such an environment. Based on this, stochastic models are proposed, formulated as additive perturbations of a limited set of measured acoustic responses and aimed at generating several additional responses with characteristics similar to those measured. In particular, we consider models with independent and identically distributed (IID) samples as well as models encompassing correlated samples. The performance estimated with the acoustic responses generated by the proposed models is assessed in terms of the acoustic contrast (AC) and compared with that estimated with acoustic responses directly measured in the vehicle. Moreover, to account for the sound perception of the human ear, the short-time objective intelligibility (STOI) is also analyzed by means of numerical simulations. The results show that with just a few measurements and simple adjustments to the few parameters of the proposed stochastic models, it is possible to obtain a good prediction of AC without the need for numerous repeated measurements. Furthermore, the listening tests and the predicted STOI performance also highlight that the proposed correlated models are able to generate realistic acoustic responses.