Compression of speech residual signal using embedded zero-tree wavelet coding

Redundancy in a speech signal is removed by passing it through an LPC analysis filter. The filter output, termed the residual error, needs to be compressed for efficient speech coding. In this paper, compression of a speech residual signal using the embedded zero-tree wavelet (EZW) coding is presented. The method does not involve any training or large database required in CELP and VSELP that are most widely used in speech compression. Contrary to other research findings, it was found empirically that the simple Haar wavelet performs as well as or better than other wavelets of higher number of vanishing moments for residual signal compression.