Production models as a structural basis for automatic speech recognition

We postulate in this paper that highly structured speech production models will have much to contribute to the ultimate success of speech recognition in view of the weaknesses of the theoretical foundation underpinning current technology. These weaknesses are analyzed in terms of phonological modeling and of phonetic-interface modeling. We present two probabilistic speech recognition models with the structure designed based on approximations to human speech production mechanisms, and conclude by suggesting that many of the advantages to be gained from interaction between speech production and speech recognition communities will develop from integrating production models with the probabilistic analysis-by-synthesis strategy currently used by the technology community. 0 1997 Elsevier Science B.V. R&urn6 Dans cet article, nous suggerons clue des modules de production de la parole fortement structures pourront contribuer significativement ?I la r&site future des mod&les de reconnaissance automatique de la parole, limit& en ce moment par les faiblesses de la base theorique de la technologie actuelle. Nous analysons ces faiblesses au niveau des modbles phonologiques et des modeles phonetiques, et presentons deux modules statistiques de reconnaissance de la parole bases sur des approximations des mCcanismes de production de la parole. Nous suggerons en conclusion que l'interaction entre les domaines de la production et de la reconnaissance de la parole peut Ctre particulierement efficace si l'on integre les modules de production dans la strategic d'analyse-synthbse probabiliste, utilisee deja depuis longtemps en reconnaissance de la parole. 0 1997 Elsevier Science B.V.