Linear prediction algorithms for lossless audio data compression

Abstract

The paper considers the use of such linear interpolation algorithms as LPC, FLPC, and Wise-LPC in the lossless audio data compression. In addition to the interpolation methods, the problems of best coding and optimal sampling window selection are investigated. The Wise-LPC algorithm is shown to allow a 1–5% improvement of audio signal compression against conventional LPC and FLPC approaches. The prediction error has a Laplace distribution, its variance decreasing smoothly and reaching “saturation” with the growing window width.