Position-based keyframe selection for human motion animation

This paper proposes a method for key-frame selection of captured motion data. Motion capture systems have been widely used in movies, games and human motion analysis. Most of the previous methods make use of the rotation angles directly and measure a cost of the rotation curves for selection of keyframes. One drawback of these methods is that they do not directly control the positions of the joints in the 3D space. Our method proposes a position-based keyframe detection scheme. In our framework, frames are decimated one by one with measuring the positions of all the joints. We make use of the cost that is the sum of all the joint position differences. Then, a frame with the lowest cost is decimated after measuring those costs about frames. We demonstrate it in experimental section by several typical motions.