Embodied Model of Speech Production

Speech production is a well rehearsed, skilled, sequential activity embodied in a complex physical system—our vocal tract. As with any form of skilled ac-tion, mastery of speech involves learning how to control and use this system in an efficient way within the constraints imposed by functional requirements, i.e. communication. There are many models which successfully and in great detail address kine-matic properties of speech production (e.g. [4, 1, 2]), but fail to take into account the generic principals underlying dynamics of motor organization like energetic efficiency, articulatory ease (possibly the same thing) and communicative efficacy ([3]). We present an early form of a developing model in which many of the com-plexities associated with physical modeling of the vocal tract are finessed, while many basic principles relevant to sequencing and coordination in real time with real masses are respected. Our abstract model consists of three simple point-mass pendula hung on massless rods from a common point on a plane. Each pendulu