Lossless audio coding aims at achieving the lowest possible bitrate for transmission or storage of audio without any loss of information. This is usually done by first removing redundancy from the audio signal, and then applying entropy coding to the residual signal. Linear prediction (LP), when applied to monophonic signals, is a very effective way to remove redundancy. It produces minimum-phase predictors that are efficiently compressed by combining vector quantization with a meaningful representation of the LP coefficients (such as the LSFs). When applied to stereo signals however, joint channel prediction often produces non-minimum-phase predictors, whose quantization requires a high bit rate and poses stability problems. In this paper, we show that backward estimation of the LP coefficients (where those are estimated on the past decoded signal) solves most of the problems associated with the use of joint channel prediction in a lossless audio coder.
(See document for exact affiliation information.)
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member Join the AES. If you need to check your member status, login to the Member Portal.
