Human Motion Synthesis with Optimization‐based Graphs

摘要: Continuous constrained optimization is a powerful tool for synthesizing novel human motion segments that are short. Graph-based synthesis methods such as graphs and move trees popular ways to synthesize long motions by playing back sequence of existing segments. However, only support transitions between similar frames, the end one segment start another. In this paper, we introduce an optimization-based graph combines continuous with graph-based synthesis. The used create vast number complex realistic-looking in graph. can then be non-trivial example allow character switch its behavior abruptly while retaining naturalness. We also propose build semi-autonomously requiring user classify generated acceptable or not explicitly minimizing amount required classifications. This process guarantees quality consistency at cost limited involvement.