Geometrical parameter combinations that correlate with early interaural cross-correlation coefficients in a performance hall.

Abstract

The previous binaural data of the authors measured inside two multi-purpose performance halls are re-analyzed using regression in this study. It is done in an attempt to establish a framework that can improve the prediction of early interaural cross-correlation coefficients (IACCs), but with as little measurement effort and parameters as possible. The results show that regression models consist of linear combinations of polynomials of geometrical parameters, when used together with the measurement schemes suggested previously by the authors, are sufficient for predicting the IACCs to within engineering tolerance. The predictions are better than those obtained previously by the neural network approach of the authors. The relative importance of the geometrical parameters in the prediction of IACCs is also investigated.