Developing Physically-Based, Dynamic Vocal Tract Models Using ArtiSynth

January 24, 2007
Sidney Fels and John E. Lloyd | Dept of Computer Science, University of British Columbia, Vancouver, Canada

We describe the process of using ArtiSynth, a 3D biomechanical simulation platform, to build models of the vocal tract and upper airway which are capable of simulating speech sounds. ArtiSynth allows mass-spring, finite element, and rigid body models of anatomical components (such as the face, jaw, tongue, and pharyngeal wall) to be connected to various acoustical models (including source filter and airflow models) to create an integrated model capable of sound production. The system is implemented in Java, and provides a class API for model creation, along with a graphical interface that permits the editing of models and their properties. Dynamical simulation can be interactively controlled through a “timeline” interface that allows online adjustment of model inputs and logging of selected model outputs. ArtiSynth’s modeling capabilities, in combination with its interactive interface, allow for new ways to explore the complexities of articulatory speech synthesis, chewing and swallowing and dysfunction of the upper airway.

Speaker Details

Sidney Fels (Ph.D, Toronto, 1994, P.Eng.) is an Associate Professor at the University of British Columbia in the department of Electrical & Computer Engineering. He was recognized as a Distinguished University Scholar at UBC since 2004. He was a visiting researcher at ATR Media Integration & Communications Research Laboratories in Kyoto, Japan from1996 to 1997. He also worked at Virtual Technologies Inc. in Palo Alto, CA developing the GesturePlus™ system and the CyberServer™ in 1995. He is internationally known for his work in human-computer interaction, neural networks, intelligent agents, new interfaces for musical expression and interactive arts with over 100 scholarly publications and exhibitions. Sidney is one of the principal investigators of the Institute of Computing, Information and Cognitive Systems through his authoring a CFI grant to create a new $22.1M facility to house interdisciplinary research.

John Lloyd received a Ph.D. from McGill University in 1995, before which he worked extensively on robotic programming and control systems. He is a principal author of the Robot Control C Library (RCCL), a software API for controlling robots directly from computer workstations. He has served as a consultant to several industry and government organizations, including General Electric and the NASA Jet Propulsion Laboratory, where he was involved in the development of space-based telerobotic systems. At the University of British Columbia (UBC) since 1995, he has done research and development work in simulation-based robotic programming, reality-based physical modeling, and contact simulation. Currently, he is the principal software developer for the ArtiSynth biomechanical modeling and speech synthesis project.