Three-dimensional motion analysis - an exploratory study. Part 2: reproducibility of facial movement

To investigate the reproducibility of using spoken word as a measure of facial movement.

Experimental study.

Department of Dental Health and Biological Sciences, University Dental Hospital, Cardiff. 22 normal subjects.

Subjects were asked to say the word 'puppy' in a normal, relaxed manner twice within a 10-second time interval. The sequence was recorded using a non-invasive, three-dimensional motion analysis image capture system (3DMDface Dynamic System) at 48 frames per second.

Corresponding frames between the two utterances were aligned three-dimensionally using best fit on non-moveable points on the upper half of the face. Reproducibility was measured as the percentage point deviation between +/- 0.5 mm between two corresponding frames.

Mean intrasessional reproducibility (SD) for the group was recorded at 86.2% (5.8). The reproducibility ranged from a minimum of 66.8% to a maximum of 97.5%. When the utterance was split into its two separate viseme segments (/p//u//p/ and /p//y/), the second part of the utterance was seen to be more reproducible than the first. The male group were more reproducible than the female group.

Intrasessional reproducibility of the utterance 'puppy' shows high intra- and intersubject variability in this group of normal subjects and therefore further research needs to be conducted before being able to confidently use this word as a reproducible measure of facial movement.