Audio Signal Processing and Coding

11.2 LOSSLESS AUDIO CODING (L²AC)

LAC schemes employ psychoacoustic principles in conjunction with time-frequency mapping techniques to eliminate perceptually irrelevant information. Hence, the encoded signal is not be an exact replica of the input audio and might contain some artifacts. Lossless coding schemes (L²AC) obtain bit-exact compression by eliminating the statistical dependencies associated with the signal via prediction techniques. Several prediction modeling methods for L²AC have been proposed, namely, FIR/IIR prediction [Robi94] [Crav96] [Crav97], polynomial approximation/curve fitting methods [Robi94] [Hans98b], transform coding [Pura97] [Geig02], high-order context modeling [Qiu01], backward adaptive prediction method [Angu97], subband linear prediction [Giur98], and set partitioning [Raad02]. L²AC algorithms are broadly classified into two categories, namely, prediction-based and transform-based.

Prediction-based L²AC algorithms employ linear predictive (LP) modeling or some form of polynomial approximation to remove redundancies in the waveform. The SHORTEN, the DVD, the MUSICompress, and the C-LPAC use LP analysis, while the AudioPaK is based on curve-fitting methods. Transform-based coding schemes employ specific transforms to decorrelate samples within a frame. The LTAC scheme proposed by Purat et al. and the IntMDCT scheme introduced by Geiger et al. are two L²AC schemes that use transform coding.

In all the aforementioned L²AC methods, the idea is to obtain ...