This is a master thesis at KTH Royal Institute of Technology
In this project, the human motion generated by speech audio would be further re-targeted on a humanoid robot NAO, which is shown above. The re-targeting procedure would be divided into two steps:
- Remap static human gestures to a NAO robot, so that the robot can mimic the gestures;
- Post-processing the motion sequence to smoothen the gesticulation and fit the physical constraints of the robot, such as maximum attainable velocity of joints.