Development of a model for generating synthetic animated lip shapes
Allen A. Montgomery · The Journal of the Acoustical Society of America · 1980
Progress is reported on a project aimed at developing the capability to synthesize realistic animations of the visible aspects of speech for the purpose of studying lip reading. An intelligent graphics system's CRT is programmed to display animated images through presenting sequences of static lip shapes. Previously it had been demonstrated that hand-copied, frame-by-frame tracings of talkers on videotape were intelligible to lip readers when redisplayed as sequences of 130-vector frames at 30 fps, and further that linear interpolation between initial and final tracings produced smoother and equally intelligible stimuli. The current model for synthesizing CVCVC stimuli from a limited set of primative shapes is described, and includes an approximation to forward and backward coarticulation and nonlinear interpolation between derived shapes. The results of lip-reading intelligibility testing are presented, the strengths and weaknesses of the model are discussed and comparisons to other possible generation stratatgies are made. [Work-supported by Clinical Investigation Service, WRAMC.]