Sound Quality Assessment Using Auditory Models

摘要

Much effort has been spent in the field of objective quality assessment and there are excellent measures for specific types of distortion. However, generalization to a wide range of data sets including various types of distortions is still a concern. This study assesses to what extent auditory model-based objective measures might improve the prediction of quality ratings by normal-hearing subjects for different types of distortion. Three intrusive sound quality assessment models, based on and including the model of Huber and Kollmeier [PEMO-Q; IEEE TASLP 14(6), 1902-1911 (2006)] were evaluated and compared to other existing quality models using three different databases. In the proposed CASP-Q model, the auditory front-end of PEMO-Q was replaced by the recent model by Jepsen et al. [CASP; JASA 124(1), 422-38 (2008)] offering non-linear auditory filters. The results revealed that the suggested quality models were able to predict a high range of different distortions and performed best compared to other state-of-the-art quality measures. In general, quality measures employing valid auditory models generalized best across different types of distortions.