Design of novel piecewise uniform scalar quantizer for Gaussian memoryless source

Abstract

[1] In this paper we propose a novel piecewise uniform scalar quantizer model designed for a Gaussian memoryless source. The segment thresholds of this quantizer model are equidistant while the number of reproduction levels inside segments is determined by optimizing the inner distortion and under the constraint of total number of reproduction levels. For the proposed quantizer model the quantization efficiency and signal to quantization noise ratio (SQNR) is studied. On the basis of the comparison with the efficiency and SQNR of uniform quantizer advantages of the proposed quantizer model are shown. Particularly, it is shown that with the proposed quantizer the correlation between the input signal and the quantization error can be eliminated more than by using the uniform quantizer as well as that the proposed quantizer provides higher level of SQNR.Citation: Velimirović, L., Z. Perić, and J. Nikolić (2012), Design of novel piecewise uniform scalar quantizer for Gaussian memoryless source, Radio Sci., 47, RS2005,

Two forward adaptive dual-mode companding scalar quantizers for Gaussian source

“…Fig. 5. shows that in the case where N ¼64, N g ¼32 and M ¼6 (R¼7 bit/sample) one can still satisfy G.712 Recommendation [17]. In other words, for the proposed forward adaptive dual-mode quantizers, we have ascertained that the bit rate of R¼7 bit/sample is the smallest one which provides high-quality quantization.…”

“…This analysis has been extended in [14], where signal to quantization noise ratio (SQNR) behavior in the wide range of variances has been analyzed for the forward adaptive PLCSQ designed for a Laplacian source according to the piecewise linear approximation to the optimal compressor function. Unlike the forward adaptive PLCSQ proposed in [14], where the number of cells has been assumed to be constant per segments and where the segments have been determined by the equidistant partition of the optimal compressor function, the number of cells per segments has been optimized in [16,17]. Similarly to [16,17], in the PLCSQ model from [15], which is based on the approximation of the first derivate of the optimal compressor function at the points in the middle of the segments, the number of cells per segments is unequal.…”

“…Unlike the forward adaptive PLCSQ proposed in [14], where the number of cells has been assumed to be constant per segments and where the segments have been determined by the equidistant partition of the optimal compressor function, the number of cells per segments has been optimized in [16,17]. Similarly to [16,17], in the PLCSQ model from [15], which is based on the approximation of the first derivate of the optimal compressor function at the points in the middle of the segments, the number of cells per segments is unequal. This resulted in SQNR increase, but in a more complex encoding/decoding procedure.…”

Application of Mean-Square Approximation for Piecewise Linear Optimal Compander Design for Gaussian Source and Gaussian Mixture Model

“…With linearized quantizer models, software and hardware realization are simplified compared to the case with an optimal compander having nonlinear compressor function [1]. In this particular field of research many different solutions for the linearization of the optimal compressor function were proposed [3], [6]- [8], [10]. The support region of these quantizers is divided into several segments, each of which contains several equal in width quantization cells and uniformly distributed reproduction levels.…”

New adaptive compandor for LTE signal compression based on spline approximations

“…Both the design and practical realization of a non-uniform companding quantizer are very complex [3]. To simplify the design procedure for a non-uniform companding quantizer and achieve an easier practical realization (in both hardware and software), linearization and approximation of the optimal compressor function are implemented in [4]. There is a new class of compandor, whose design is based on an approximation of the optimal compressor function by firstdegree spline functions.…”