Downmix and coding of multichannel signals based on spatial correlation

摘要

In multichannel audio system, virtual sound source may be produced and rendered at arbitrary directions and locations so the multichannel signals are complicated. Parametric coding methods encode two or more channel signals into a downmix accompanied with spatial parameters, which can provide preferable compression ratio and considerable sound quality in multichannel audio coding. Due to the correlation between channels the signal redundancy in channels will be removed in the downmixing. However the downmixing of two uncorrelated channel signals will cause virtual sound aliasing and seriously decrease sound quality. A spatial virtual sound oriented correlation analysis method between channel signals is proposed to combine the highly correlated channels to groups for downmixing and parameters extracting. Both subjective and objective experimental results for sound quality and spatial orientation evaluation identify that the proposed dynamical grouping method for downmixing is superior to the fixed grouping method in MPEG Surround. The proposed analysis method of spatial correlation between channel signals has much lower computational complexity than the traditional direct computing of correlation and provides considerable accuracy.