To develop a physiological articulatory model for speech production, mastication and swallowing, we proposed an analysis-by-synthesis (AbS) based estimation method for investigating contributions of the tongue muscles in both exterior movement and interior deformation using observations and model simulations. The validity of the method was confirmed by comparing the estimated muscle activation to known muscle activation via model simulation using a full three dimensional physiological articulatory model. Observations were conducted to measure several simple tasks of large-scale tongue deformations using tagged magnetic resonance imaging (tMRI), and the principal strain of the deformations were analyzed using HARP MRI analysis. The muscle activation was examined based on the principal strain pattern and anatomic knowledge. According to the observation, a muscle activation pattern was designed to generate a tongue movement with extreme protrusion-retraction using the model. Muscle activation in the movement was estimated by comparing the principal strain pattern and known muscle forces. The model simulation and tMRI observation showed consistent strain patterns with each other. The study showed that the muscle activation within tongue movement is able to be estimated from local deformations using the AbS method.
[1]
J. Dang,et al.
Construction and control of a physiological articulatory model.
,
2004,
The Journal of the Acoustical Society of America.
[2]
J R Westbury,et al.
Vowel posture normalization.
,
1998,
The Journal of the Acoustical Society of America.
[3]
M. Honda,et al.
Generation of articulatory movements by using a kinematic triphone model.
,
2001,
The Journal of the Acoustical Society of America.
[4]
W S Levine,et al.
Modeling the motion of the internal tongue from tagged cine-MRI images.
,
2001,
The Journal of the Acoustical Society of America.
[5]
H. Takemoto,et al.
Morphological analyses of the human tongue musculature for three-dimensional modeling.
,
2001,
Journal of speech, language, and hearing research : JSLHR.