Abstract:

The Personal Sound Zone （PSZ） technology in automotive cabins is a spatial sound field reproduction technique that uses an array of speakers to create independent sound zones within the cabin. When designing the speaker control signals， it is essential to accurately obtain the acoustic transfer functions （ATFs） between the speakers and the control points within the targeted area. However， in practice， various interferences prevent the ideal ATFs from being obtained， resulting in degraded sound field reconstruction performance. Traditional robust control methods typically add regularization terms to the optimization problem， but these methods face challenges in selecting appropriate regularization parameters. To address the problem， a robust ACC-LD method based on a probabilistic model of transfer function errors is proposed in this paper. This approach fits the errors of the ATFs using the Gaussian mixture model， and the Expectation-Maximization （EM） algorithm is employed to train the model and obtain the distribution parameters. The errors are then incorporated into the ACC-LD optimization problem and solved by taking the expectation， thereby avoiding the parameter tuning issues inherent in regularization methods. A comparative study in the simulated automotive cabin environment shows that the proposed method outperforms traditional regularized robust control methods in terms of contrast between light and dark area as well as the consistency in the bright zones.

Key words: sound field zoning within a car cabin, robust control, error probability model