Vocal cord model to control various voices for anthropomorphic talking robot

摘要

We have developed a biomechanical vocal cord model for an anthropomorphic talking robot—the Waseda Talker Series. The model mimics the biomechanical structure of human vocal cords and is made of a soft rubber called Septon. It is able to reproduce vocal cord vibrations of the normal (modal) voice in a form similar to the human. We also developed a neural network based model for controlling the vocal cord parameters with acoustic parameters of the speech signal. The control model consists of a forward and inverse model, and a simple feedback model. We also describe voice quality control for the vocal cord model which is important for generating fluent and emotional speech. 1