Three-dimensional sound reproduction in vehicle based on data mining technique

摘要

Three-dimensional (3D) audio reproduction technique is a hot topic since MPEG proposed a proposal for 3D audio standard in 2012. The present 3D reproduction method is not available in vehicle because the loudspeaker configuration does not meet the requirements of the methods. In this paper, we develop a regression model to reproduce sound pressure in vehicle using a data mining algorithm. A big sound pressure dataset is established based on acoustic theory. We analyze this dataset and propose a stepwise regression model training and testing on this dataset. Both the Mean Square Error and Root Mean Square Error are rather small, and all the residuals such as raw residuals, Pearson Residuals, Studentized residuals, and Standardize residuals are small as well. The experiments indicate that the proposed reproduction system accurately simulates the theoretical sound reproduction system.