摘 要
一般運動學習軟體無法使用虛擬角色做即時回饋的原因，在於動畫設計師無法針對所有可能互動的情境作繪製。本研究提出以標籤配對的方式來銜接不同動作，透過模糊理論定義動作橋接影格的標籤，讓虛擬角色的動作與動作之間能透過動作橋接影格進行轉換，讓虛擬角色有多元的互動呈現，不會因為切換不同的影片片段而產生不連續的動作轉換。本研究開發一套 3D 虛擬角色，利用帶動唱的方式來教導小朋友跳舞，系統使用 Microsoft Kinect 鏡頭捕捉與分析玩家動作來判斷玩家動作是否正確，虛擬角色能夠重複示範動作的方式，對使用者的錯誤動作給予即時的指導回饋。
關鍵詞：角色動畫、動作合成、動作標籤、虛擬角色
Abstract
Immediate feedback is difficult for virtual characters in fitness video games because animators cannot design keyframes for all possible interactions. This study proposes fuzzy marked keyframe bridging techniques to connect different video clips. We use the fuzzy theory to define markers for bridging keyframes. These markers can be used as switch points between two movie clips and then generate the actions of the avatar. The purpose of this research is to develop the automatic generation of interactive feedback motions of avatars without discontinuous motion caused by switching between different movie clips. This study develops a 3D virtual characters video game to teach children to dance by singing. The system uses the Microsoft Kinect to capture the player's movements and postures and analyze whether the postures are correct, and then virtual characters provide immediate coaching feedback and repeat actions until the player can perform the correct movement.
Keywords: Character animation, Motion synthesis, Marked keyframe, Avatar