Control system for talking robot to replicate articulatory movement of natural speech

The ultimate goal of our study is to create a new speech production system, in which an anthropomorphic hardware talking robot is handled so as to imitate human articulatory movement. We have developed a software motion simulator, which generates control parameter sequences for the talking robot Waseda Talker No. 2 (WT-2) using the trajectories of human articulatory organs during continuous utterances measured by an electromagnetic articulograph. This paper mainly describes the motion simulator. In addition to its main function as a generator of control parameters for WT-2, the motion simulator also simulates the resultant acoustic characteristics related to WT-2's vocal tract shape at each instance during the motion. The hardware structure of WT-2 is also described briefly. This comprehensive approach will enable us to study speech production using the features of the vocal tract shape as speech motor tasks, instead of using acoustic features.