Skip to Main content Skip to Navigation

Etude des caractéristiques de la langue vietnamienne en vue de sa synthèse et de sa reconnaissance automatique. Aspects statiques et dynamiques

Viet Son Nguyen 1 
1 Speech Communication
MICA - International Research Institute MICA
Abstract : This research aims to arrive at an in-depth understanding of the static and dynamic characteristics of Vietnamese vowels (long and short) and final consonants. Measurements of the speed of formant transitions constitute a key aspect of the work. Two sets of combinations are analyzed: (i) (C1)V1C2, with C2 one of /-p, -t, -k/; and (ii) (C1)V1V2, exploring combinations of long and short vowels. The aim is to bring out specific characteristics, and to attempt to apply them in speech synthesis. Two data sets were created. The main results are as follows: (i) The Vietnamese short and long vowels within the three pairs /a, ӑ/, /ɤ, ɤ̆/, /ɔ, ɔ̆/ are acoustically close to each other in the F1-F2 plane; they are distinct in terms of overall duration, of slope of formant transitions, and of duration of the V2 semivowel within V1V2 combinations. No differences were found between short and long vowels in terms of duration of transitions in V1C2 and V1V2. (ii) Final /p, t, k/ are unreleased: they terminate on silence, without a burst. For (C1)V1C2 syllables, comparison between different consonants in C2 position (preceded by the same V1) shows significantly different dynamic acoustic characteristics (slope of formant transitions), whereas no differences were found in terms of static acoustic characteristics (duration of the transition, and formant values at the onset of formant transition). Thus, dynamic characteristics arguably constitute decisive cues to the perception of Vietnamese final consonants. (iii) Locus equations with striking linearity were obtained, with values that cluster neatly around regression lines. The places of articulation of the three final stops /p/, /t/, /k/ are clearly distinct. No effect of tone on the locus of final consonants was observed. These phonetic results are verified through the use of synthesis systems: SMART (Synthesis with a Model of Anthropomorphic Region and Tract) et DRM (Distinctive Region Model).
Complete list of metadata

Cited literature [94 references]  Display  Hide  Download
Contributor : Alexis Michaud Connect in order to contact the contributor
Submitted on : Wednesday, September 17, 2014 - 12:15:20 PM
Last modification on : Friday, May 13, 2022 - 1:54:02 PM
Long-term archiving on: : Thursday, December 18, 2014 - 10:51:02 AM


  • HAL Id : tel-01064853, version 1




Viet Son Nguyen. Etude des caractéristiques de la langue vietnamienne en vue de sa synthèse et de sa reconnaissance automatique. Aspects statiques et dynamiques. Traitement du signal et de l'image [eess.SP]. Telecom ParisTech; Institut Polytechnique de Hanoi (Hanoi University of Science and Technology, HUST), 2009. Français. ⟨tel-01064853⟩



Record views


Files downloads