著者

福井 孝太郎 新宅 英滋 下村 彰宏 榊原 菜々 石川 優馬 誉田 雅彰 高西 淳夫

出版者

The Robotics Society of Japan

雑誌

日本ロボット学会誌 (ISSN:02891824)

巻号頁・発行日

vol.26, no.7, pp.794-800, 2008-10-15

参考文献数

11

---

We have developed a vocal control method based on forward and inverse model, and new vocal cord mechanism, for anthropomorphic talking robot, Waseda Talker No.7, to produce various kinds of voices. The previous artificial vocal cord model could reproduce human-like vibration, however, the control range of the acoustic parameters are very small. We developed new mechanism could adjust the vocal cords tension effectively and control the glottal opening-closing, to broaden the control range. The control parameters of the vocal cords on WT-7 are vocal cord tension, glottal opening and air pressure, and the acoustic parameters are sound pitch, spectrum tilt and sound pressure. The relations between these parameters are complicated, and hard to modeling. The learning process consists of learning of the NN forward model and learning of the inverse model by using the forward model. In addition, the real-time auditory feedback mechanism is used to reduce the error. By this method, the acoustic parameters could be followed well the target.