Effective utilisation of JND for spatial parameters quantisation in 3D multichannel audio

摘要

The features of just noticeable difference (JND) in human spatial perception are often used to remove the perceptual redundancy of spatial parameters in 3D multichannel audio. However in previous spatial parameters quantisation schemes, JND data are not effectively utilised resulting in either perceptual distortions of spatial images, or wastes of coding bitrates. It is proposed to effectively utilise azimuthal JND to design the quantisation codebook of spatial parameters azimuths, which divides the full circle of 360° to adjacent quantisation intervals, and set the endpoints of quantisation intervals according to JND of quantisation value. With presented scheme, quantisation codebook size has decreased 13 and 57%, respectively, compared with MPEG Surround and SLQP, as well as ignorable perceptual distortion induced.