New Coding Algorithm Based on Variable-Length Codewords for Piecewise Uniform Quantizers
Volume 27, Issue 3 (2016), pp. 527–548
Pub. online: 1 January 2016
Type: Research Article
Received
1 January 2014
1 January 2014
Accepted
1 August 2015
1 August 2015
Published
1 January 2016
1 January 2016
Abstract
This paper presents a model for signal compression, which consists of a piecewise uniform quantizer and a new lossless coder. The model is designed in a general manner, i.e. for any symmetrical signal distribution; this general theory is applied to design models for Gaussian and Laplacian distributions. Rigorous mathematical derivation of the expression for the bit-rate is presented. Forward adaptation of the model is done for non-stationary signals. Theory is proved by simulations in MATLAB and by an experiment with a real speech signal. The most important advantages of the model are low complexity and good performances – it satisfies G.712 standard for the speech transmission quality with 6.18 bps (bits per sample), which is significantly smaller than 8 bps required for quantizers used in PSTN (public switched telephone network) defined with G.711 standard.