Abstract
This paper proposes a method to generate robot's behaviors representing emotions reminded from the given speech. The robot learns the relationship between words contained in speech and behaviors by using the machine learning technique. This makes the robot possible to generate its behaviors automatically that correspond to spoken words. The robot's behaviors will become more human-like and this will widen the application fields of the robot. The proposed method has several advantages, that is, the generation of robot's behaviors from the speech containing unknown words, the variety of generated behaviors even for the same speech, and the mutual estimation from speech to behavior and vice versa. The effectiveness of the proposed method has been shown through the subjective evaluation experiments of robot's behaviors generated by the proposed and previous methods.
