Biomechanical Modeling of the Tongue

While human facial animation is a hotbed of research activity, some parts of the face receive less attention than others. In particular, research tends to gloss over the tongue. Moreover, typical models of the human face do not allocate even a single degree of freedom for the tongue. Admittedly, acceptable animation is entirely possible even if the tongue is only modeled as static geometry, but a more sophisticated model would further enrich the expressiveness of virtual faces, and enhance the realism of animated speech. This project presents a simplified, but anatomically derived, biomechanical model of the human tongue. Anatomy of the Tongue The tongue's relatively simple musculature make it an ideal candidate for modeling. The muscles of the tongue can be divided into two distinct groups: extrinsic, and intrinsic. The extrinsic anchor the tongue to bone structures in the head and neck, while the intrinsic are contained completely within the tongue itself. In general, the extrinsic muscles of the tongue control its position and orientation, while the intrinsic control its shape. Much like muscles of the neck, the muscles of the tongue play redundant roles making control of the tongue somewhat challenging proposition. Figure 1. Extrinsic Muscles of the Tongue [5] Figure 2. Intrinsic Muscles of the Tonge [5] Related Work The inspiration for this project comes from two existing works. The first is a biomechanical model of the neck [2]. Similar techniques might be applicable when modeling and controlling other parts of the body, including the tongue. That said, while the muscles of the neck control only a small number of unambiguous degrees of freedom, the deformable tongue's degrees of freedom are somewhat subjective. For example, If we only consider the position of the tip of the tongue, the system is decidedly over-actuated given the number of muscles in the tongue. On the other hand, if we consider the position of every element of the tongue, then the system is severely under-actuated. In this respect, the tongue is more like the second work that inspired us: A biomechanical fish [3]. The fish's individual degrees of freedom alone fail to convey the higher level control goals we have in mind (swimming, turning, etc...) Ultimately, the work shows this can be overcome as long as we can represent our goal as an objective function. Such higher level goals for the tongue might include maximizing its length, positioning the tip of the tongue at a specific location, etc... In any case, we need a model the tongue before we can tackle any control issues. Modeling of the human tongue is admittedly more well travelled than we initially expected. [1] In particular presents a highly detailed and sophisticated tongue model. The purpose of this model, however, is the study of speech motor control, and consequently it is somewhat overcomplicated for most animation applications. Thus, we decided to press on and see if a simpler model could produce a result acceptable result. Implementation We chose to model the physics of our project with Bullet [6] an open source physics library. The inert tissue of the tongue is modeled with generic soft bodies provided by Bullet. The muscle actuators of the tongue are implemented as springs with controllable rest lengths. The spring muscles are then affixed to the mesh of the soft body, and to static positions outside the tongue. The connection points and the parameters of the spring-muscles were determined by a combination of experimentation and consultation of the tongue anatomy[4],[5] The polygonal mesh of the tongue a one taken from an existing finite element model of a real human tongue [4]. Figure 3. The muscles of our simplified model (side view) A video of our model in action can be viewed at the author's website: http://www.cs.ucla.edu/~alexalex/TonguePage.html Conclusion We have presented an anatomically derived biomechanical model of the human tongue using spring dampers to model actuator muscles and soft bodies to model the inert tissue. While our model is functional, it is not yet especially believable. Issues include general wobbliness, insufficient flexibility of the inert tissue, and an inability to precisely reshape tongue tip. To improve our model we will need to further tune the parameters of the muscles and consider an alternative model for the tongue's inert tissue. Future Work Though our result produced an acceptable real time result, a superior result would likely be attained by directly retooling the model from [1], and integrating it into [2]. Furthermore, it would be essential abstract away theredundancy of actuators in the tongue and provide an intuitive, yet flexible way of providing pose control foranimators. Techniques for achieving this have already been demonstrated in [2] (using neural nets) and [3] (usingsimulated annealing to maximize objective functions). A natural application of the former methodology would bechoosing the inputs to be the position of the tongue tip. The latter technique in particular could be used to devisethe effective motion for lapping up virtual water. Also important would be integrating the tongue model with the greater head model and ensuring properinteraction of the tongue with the manible, teeth and roof of the mouth.References[1] J. Gérard, R. Wilhelms-Tricarico, P. Perrier, Y. Payan, "A 3D dynamical biomechanical tongue model tostudy speech motor control," TIMC-IMAG Laboratory 2003: http://www-timc.imag.fr/IMG/pdf/Gerard_et_al.pdf [2] S. Lee and D. Terzopoulos, "Heads Up! Biomechanical Modeling and Neuromuscular Control of theNeck," Proceedings of the ACM SIGGRAPH 2006 Conference [3] R. Grzeszczuk, D. Terzopoulos, "Automated learning of muscle-actuated locomotion through controlabstraction," Proceeding of the ACM SIGGRAPH'95 Conference, Los Angeles, CA, August, 1995, in ComputerGraphics Proceedings, Annual Conference Series, 1995, 63–70. [4] Current state of the tongue model: http://webpages.charter.net/reinerwt/themodel.htm [5] Muscles of tongue: http://en.wikipedia.org/wiki/Muscles_of_tongue [6] Bullet Physics: http://www.bulletphysics.com