scispace - formally typeset
Search or ask a question

Showing papers by "Philippe Gournay published in 2000"


Journal ArticleDOI
TL;DR: The hsx technique for coding at 1200 bps and a new segmental method with automatically derived units for very low bit rate coding are presented in details.
Abstract: Cet article presente les principales techniques de codage de parole a bas et tres bas debits, de 50 bit/s a 4 000 bit/s. Puis il presente en detail la methode hsx pour le codage a 1200 bit/s et une nouvelle approche segmentale utilisant des unites acoustiques obtenues de maniere non supervisee pour des debits inferieurs a 400 bit/s.

10 citations


01 Jan 2000
TL;DR: In this paper, the parametres prosodiques of a codeur de parole a tres bas debit (400 bits/s) combinant reconnaissance and synthese vocale are decrited.
Abstract: Cet article decrit la maniere dont ont ete codes les parametres prosodiques d’un codeur de parole a tres bas debit (400 bits/s) combinant reconnaissance et synthese vocale. Ce codeur utilise des segments de parole de longueurs variables. Les classes d’unites acoustiques sont determinees de facon automatique. Le codage de la prosodie repose principalement sur l’utilisation de la segmentation obtenue lors du codage de l’enveloppe spectrale et exploite les caracteristiques prosodiques communes des segments codes par une meme unite representante.

7 citations


01 Jan 2000
TL;DR: This paper presents a study of the influence of a singlesensor noise pre-processing method applied before coding on the performance of a parametric low bit rate speech coder, and proposes two new noisepre-processing methods based on the combination of two algorithms.
Abstract: This paper presents a study of the influence of a singlesensor noise pre-processing method (i.e. applied before coding) on the performance of a parametric low bit rate speech coder. The speech coder we consider estimates four parameters: the fundamental frequency, a voicing information, a set of linear prediction coefficients, and the energy. We first study the influence of background noise and noise pre-processing on those parameters and on the global output intelligibility. Then, we propose two new noise pre-processing methods based on the combination of two algorithms. Finally, we present some tests results that underline the advantages gained from using a combined pre-processing technique.

3 citations