Developing synchronous source generation for PC-controlled Umeda-Teranishi’s vocal-tract model

Vowel production has been investigated by Umeda and Teranishi using a physical model of the human vocal tract with a set of movable bars. This model was later combined with actuators to enable PC control of the vocal-tract configuration. The system primarily controlled the positions of the bars to enable the production of dynamic aspects of speech, such as uttering phrases or sentences, while the source was pre-generated in advance. However, a limitation remained: the source was not always played back in perfect synchrony. To address this issue, we took the following two steps. First, we developed a computer program with a GUI that enabled users to specify any fundamental frequency contour by moving a pointing device, thereby creating arbitrary intonation. Second, we modified a mechanical component so that one of the actuators controls the fundamental frequency contour via the pointing device. As a result, we confirmed that articulator movements and intonation were perfectly synchronized during phrase production. Work supported by JSPS KAKENHI Grant No. 24K06423.