A mass-spring tongue model with efficient collision detection and response during speech

Tongue modeling and animation are important for multimodal observation and analysis of speech. Specially, combining the tongue animation with other speech related techniques can improve the effect of spoken language learning. In this paper, a mass spring based real time tongue animation system with the functions of collision detection and response to stimulus is presented. For tongue modeling, we use the magnetic resonance imaging (MRI) data of tongue to generate a tetrahedral mesh which is used to build a mass spring tongue model capable of representing various tongue shapes during speech. For tongue animation, the electromagnetic articulography (EMA) is used to capture the various kinds of movements of tongue. The obtained EMA data is then used to mimic the dynamic behavior of tongue. Besides, an efficient collision detection algorithm is introduced for detecting the collisions between the tongue and the teeth or the jaws during speech. The collision detection results can be used to control the local deformations of the tongue after collisions. The experimental results show that our method is effective for generating a realistic tongue animation which include collisions handling during speech, especially for Chinese speech.