Claude Lamblin

Design and description of CS-ACELP: a toll quality 8 kb/s speech coder

This paper describes the 8 kb/s speech coding algorithm G.729 which has been standardized by ITU-T. The algorithm is based on a conjugate-structure algebraic CELP (CS-ACELP) coding technique and uses 10 ms speech frames. The codec delivers toll-quality speech (equivalent to 32 kb/s ADPCM) for most operating conditions. This paper describes the coder structure in detail and discusses the reasons behind certain design choices. A 16-b fixed-point version has been developed as part of Recommendation G.729 and a summary of the subjective test results based on a real-time implementation of this version are presented.

Fast CELP coding based on the Barnes-Wall lattice in 16 dimensions

The authors present an algebraic code-excited linear prediction (CELP) speech coder where the innovation codebook comes from the first spherical code of the Barnes-Wall lattice in 16 dimensions. Novel fast optimal algorithms for finding the best sequence in this Barnes-Wall shell innovation codebook are described. This algebraic codebook makes it possible to design a CELP coder at 9.6 kb/s with good quality and still implementable on a current digital-signal-processing chip.<<ETX>>

Algorithme de quantification vectorielle sphérique à partir du réseau de Gosset d’ordre 8

RésuméLes connaissances sur les réseaux réguliers de points trouvent leurs applications aussi bien dans la conception d’un quantificateur de formes d’ ondes que dans la sélection d’un jeu de signaux pour la modulation. Dans le domaine du codage de la parole, une nouvelle technique récemment proposée [1], la quantification vectorielle algébrique sphérique, utilise des sous-ensembles sphériques extraits de réseaux réguliers de points. L’article concerne la quantification vectorielle algébrique sphérique basée sur le réseau de Gosset de dimension huit. Un algorithme optimal rapide est décrit en détail qui permet de déterminer le plus proche vecteur parmi une collection finie sphérique du réseau de Gosset et de déterminer le rang de ce plus proche voisin dans la collection. Les performances de ce quantificateur pour le bruit blanc gaussien sont comparées aux limites prédites par la théorie de l’information. La comparaison des approches algébrique et statistique amène à envisager la conception de quantificateurs hybrides intégrant ces deux techniques.AbstractRegular point lattices find applications both for the design of waveform quantizers and the design of modulation signal sets. Algebraic spherical vector quantization is a technique recently proposed [1] for speech coding applications. It makes use of spherical subsets from regular point lattices. The article concerns itself with algebraic spherical vector quantization based on the so-called Gosset lattice in eight dimensions. An optimal algorithm is detailed which allows an efficient nearest neighbour search within a Gosset spherical set and computes the rank of this nearest neighbour within the set. The performances of such quantizer for the white Gaussian process are given and compared to the information theory limits. Comparing the merits of the statistical and the algebraic approaches to vector quantization leads to the design of hybrid structures which can combine these two techniques.