Abstract:

We present a mechanical vocal cord model aiming for a talking robot, WT-5 (Waseda Talker No. 5). Unlike a musical reed which has been used in conventional mechanical speech synthesizer, the vocal cord model is formed to mimic the human's vocal cord in the shape and the biological structure. It is made of a thermoplastic rubber, Septonh (Kuraray Co. Ltd.) of which the elasticity like a human's, and has 3-DOF mechanisms which is similar to the human structure. 1-DOF link mechanism could change the pitch by stretching the length of the vocal cords. The 2-DOF arm mechanism is used to mimic the abduction and adduction of a human arytenoid cartilage. The vocal cord model was excited by air flow exhausted from a mechanical lung model. The vibration pattern was observed by a high-speed camera, and the glottal volume velocity and the sound pressure were recorded by a mask-type wire screen pneumotachograph and a microphone. It was shown that the lower and upper edges of the vocal cords could vibrate in a different phase and the sound spectrum was similar to the source spectrum of human speech.