Piecewise uniform scalar quantization technique for Laplacian source with application in speech coding

This paper propose two novel methods for design of piecewise uniform scalar quantizers that differ in the fact how the support region is partitioned into the segments as well as in the fact how the reproduction level distribution per segments is carried out. By designing the proposed quantizers for the assummed Laplacian memoryless source and by studing the SQNR behaviour, the conclusion about the SQNR degradation due to the introduction of linearization is carried out.