Cyclification of Human Motion for Animation Synthesis

Abstract

In this paper, new techniques are introduced for matching cyclic motion boundaries by using simple animation processing algorithms based on some observed characteristics from different families of cyclic motions. Matching the cycle boundaries is one of the essential stages in generating a stream of animation of cyclic human movements. If the cycle boundaries are not matched, noticeable artefacts and flickers will appear regardless of how realistic the available motion cycle is. The proposed techniques are developed to modify the motion cycle even in the case, where only one cycle is provided. These techniques do not only match the cycle boundaries but also correct other parts of the cycle and motion constraints. The techniques are computationally efficient as they work directly on the animation data (without conversion to intermediate representations such as spline or time-frequency transforms). Moreover, many aspects of animation (including joint limits and important motion constraints) are preserved as the modification is guided by the original motions. A simple framework for automatic detection of gait phases timing is also introduced.