Pose Generation System Expressing Feelings and State

摘要

A pose generation system expressing feelings and the state is proposed in this paper. The most important contribution of this paper is proposal of a pose generation system with high expression ability. Generally, gestures are composed of pose, facial expression and the movement. This paper treats pose and facial expression. The proposed system generates a pose in accord with the image that the user wants to make, and it outputs avatars with the poses. The operation and the usage of the system are very simple, and high quality avatars with poses can be obtained. First, a user chooses onomatopoeias which are close to the image of the pose that the user wants to create. Onomatopoeia is a word, or occasionally, a grouping of words, with a sound imitating the sound it is describing, such as “buzz” for a bee and “moo” for a cow. Then the system determines the initial values of the parameters to constitute a pose from the chosen onomatopoeias. Various poses can be generated by interactive genetic operations based on the initial values, and they are displayed to the user. Then the user selects and evaluates his/her favorite poses, and the system learns preference of the user based on the evaluations. In this way, interactive genetic operations are performed and more and more suitable poses are generated. Furthermore, since addition of facial expression is possible, various feeling/state expressions are realized. In addition, the poses are expressed in a 3D space, a viewpoint can be changed freely on the screen. Effectiveness of the proposed system is confirmed through evaluation experiments.