An extensive audio-visual corpus of transitions between French phonemes was recorded (signal, plus front and profile views) from a speaker with high dynamics and symmetry in lip displacements. Nine repetitions of 14 vowels and 6 consonants in various combinations were analysed, and 3300 video fields were labeled on the centres of the acoustic wave-forms. The vermilion zone of the speaker's lips was carefully made up in an intense blue, and a Chroma-key transformed the lips area to a saturated black color in real-time. Each selected video field was digitized on PC-based "visual speech" workstation. The internal and external contours of the vermilion zone were automatically extracted and stored. A program measured various surfaces from these patterns (upper, lower, and inter-lip areas) and various distances between reference (front and profile) lines and anatomical points. 17 parameters were ultimately measured and computed from 774 selected video fields. Clustering and discriminant analysis allowed us to isolate 19 (± 2) lip-jaw shapes that constitute the French speaking "labial space". These visemes are here described.
Cite as: Benoît, C., Lallouache, T., Mohamedi, T., Tseva, A., Abry, C. (1990) Nineteen (±two) French visemes for visual speech synthesis. Proc. First ESCA Workshop on Speech Synthesis (SSW 1), 253-256
@inproceedings{benoit90_ssw,
author={Christian Benoît and T. Lallouache and T. Mohamedi and A. Tseva and Christian Abry},
title={{Nineteen (±two) French visemes for visual speech synthesis}},
year=1990,
booktitle={Proc. First ESCA Workshop on Speech Synthesis (SSW 1)},
pages={253--256}
}