This paper describes the development of the objective method for the prediction of the Basic Audio Quality (BAQ) of bandlimited or down-mixed surround audio recordings. A number of physical parameters, including interaural cross-correlation coefficient and spectral descriptors, were extracted from the recordings and used in a linear regression model to predict BAQ scores obtained from listening tests. The results showed a high correlation between the predicted scores and those obtained from the listening test, with the average error of prediction being smaller than 10%. Although the method was originally developed for 5-channel surround recordings, after some modifications it can be upgraded to any number of audio channels.
展开▼
