Speech spectral quantizers for wideband speech coding

Abstract

In this treatise a range of Line Spectrum Frequency (LSF) Vector Quantization (VQ) schemes were studied comparatively, which were designed for wideband speech codecs. Both predictive arrangements and memoryless schemes were investigated. Specifically, both memoryless Split Vector Quantization (SVQ) and Classified Vector Quantization (CVQ) were studied. These techniques exhibit a low complexity and high channel error resilience, but require high bit rates for maintaining high speech quality. By contrast, Predictive Vector Quantizers (PVQ) offer an enhanced Spectral Distortion (SD) performance, although they are sensitive to channel error propagation. It is shown that the family of so—called Safety—Net Vector Quantization (SNVQ) schemes offers a good design compromise, providing an extension to memory—based PVQ, and thereby improving the performances both over noisy and noiseless channels.

Sequential Split Vector Quantization of LSF Parameters using Conditional Pdf

“…However, the statistical stability and relative improvement need to be consistent. 2The uniform bit allocation is necessary to keep the search complexity minimum ( [4], [12]). If 24 bits/vector is available, then 8 bits are allocated to each of the three sub-vectors.…”

“…There are many different LSF VQ methods reported in the literature; in particular, several structured VQs [2] have been proposed for telephone-band speech and further extended to wide-band speech [12]. Some of the recent techniques are parametric VQ [14], HMM based recursive quantizer [15] and two stage transform vector quantization [17].…”

Low-Complexity Wideband LSF Quantization Using Algebraic Trellis VQ

“…The listening tests of Guibé et al [19], have shown that the existing requirements for transparency in narrowband LPC coding also apply to the wideband case.…”

A comparison of LSF and ISP representations for wideband LPC parameter coding using the switched split vector quantiser

“…Transparent results were reported by Biundo et al [4] for a four and five part split vector quantiser at 45 bits/frame. Because successive LSF frames are highly correlated [7], better quantisation can be achieved by exploiting the interframe correlation. Ubale and Gersho [20] used a seven-stage tree-searched multistage vector quantiser [10] with a moving average (MA) predictor at 28 bits/frame, while Biundo et al [4] reported transparent results using an MA predictive split-multistage vector quantiser (S-MSVQ) at 42 bits/frame.…”

“…Ubale and Gersho [20] used a seven-stage tree-searched multistage vector quantiser [10] with a moving average (MA) predictor at 28 bits/frame, while Biundo et al [4] reported transparent results using an MA predictive split-multistage vector quantiser (S-MSVQ) at 42 bits/frame. Guibé et al [7] achieved transparent coding using a safety-net vector quantiser at 38 bits/frame, while the Adaptive Multi-Rate wideband (AMR-WB) speech codec [2,1] uses an S-MSVQ with MA predictor at 46 bits/frame. Other quantisation schemes recently reported include the predictive Trellis-coded quantiser [15], the HMM-based recursive quantiser [6], and the multi-frame GMM-based block quantiser [18], which achieve a spectral distortion of 1 dB at 34, 40, and 37 bits/frame, respectively.…”