Prediction of Perceived Quality in Multi-Channel Audio Compression Coding Systems

Objective quality assessment methods, such as described in ITU-R Recommendation BS.1387-1 [1], have been widely used for evaluation of audio coding systems. However, even though many different multi-channel audio compression coding systems are being developed, most current quality assessment methods only predict results for monaural or stereo signals. In this paper, a prediction method is introduced that can be used for the objective quality assessment for multi-channel audio compression coding systems. The method introduces two novel variables, interaural level difference distortion (ILD distortion) and interaural cross-correlation coefficient distortion (IACC distortion) to predict degradations in spatial quality. Simultaneously, five Model Output Variables proposed in ITU-R BS.1387-1 are selectively extracted from binaural signals that are synthesized using binaural room transfer functions. The prediction model is trained and verified using results from subjective listening tests of multi-channel audio compression coding systems that were performed by participants in MPEG audio group. This new model, using the two interaural and five non-spatial statistics, shows encouraging results in prediction perceived quality.