METHOD AND APPARATUS FOR AUDIO OBJECT CODING BASED ON INFORMED SOURCE SEPARATION

摘要

A kind of source of the composition in mixture in order to indicate and restore to be present in sound is separated using well-informed sources and is reacted. Particularly, a universal spectral model (USM) is used to obtain the rare time activated matrix for the individual audio source in the mixture of sound. The index of non-zero groups in time activated matrix is encoded as side information into a bit stream. The nonzero coefficient of time activated matrix can also be encoded into bit stream. In decoder side, when the coefficient of time activated matrix is included in bit stream, matrix can be decoded from bit stream. Otherwise, time activated matrix can be by the mixture of sound, and the non-zero indices being included in bit stream and USM models are estimated. In view of time activated matrix, constituent audio source can mixture based on sound and USM models be recovered.