Three‐dimensional mechanical tongue and vocal cord models for human speech production.

Abstract

We have developed talking robot Waseda Talker series, which mimicked human speech production mechanism. This robot is consisted of mechanical vocal cords, tongue, lips, and nasal cavity. The current version, WT‐7R (Waseda Talker No. 7 Refined), has 16 DOF (7 DOF in tongue, 1 DOF in jaw, 4 DOF in lips, 1 DOF in velum, 2 DOF in vocal cords, and 1 DOF in lung), having same vocal tract length as average adult male. The mechanical vocal cord model is made of styrene based thermoplastic elastomer “Septon” and the shape is mimicking vocal folds of human with three dimensions. The vocal code model has pitch control mechanism by adjusting the cord tension and glottal opening‐closing mechanism, and it could reproduce the vocal cord vibration of the vocal fry (double pitched) and the breathy voice (unclose in one cycle) as well as the modal voice with variable pitch. The three‐dimensional tongue model is constructed by the rigid link mechanism covered by the Septon soft rubber. The tongue model could be controlled so as to configure the tongue shape in a three‐dimensional way. We will describe the details of the mechanism of talking robot by showing some demonstration videos.