- 著者
-
向井 智彦
栗山 繁
- 出版者
- ACM
- 雑誌
- Proceedings of ACM SIGGRAPH 2005 (ACM Transactions on Graphics Vol.24 Issue 3, July 2005)
- 巻号頁・発行日
- vol.24, no.3, pp.1062-1070, 2005-07
- 被引用文献数
-
112
A common motion interpolation technique for realistic human animationis to blend similar motion samples with weighting functionswhose parameters are embedded in an abstract space. Existingmethods, however, are insensitive to statistical properties, suchas correlations between motions. In addition, they lack the capabilityto quantitatively evaluate the reliability of synthesized motions.This paper proposes a method that treats motion interpolationsas statistical predictions of missing data in an arbitrarily definableparametric space. A practical technique of geostatistics, calleduniversal kriging, is then introduced for statistically estimating thecorrelations between the dissimilarity of motions and the distancein the parametric space. Our method statistically optimizes interpolationkernels for given parameters at each frame, using a posedistance metric to efficiently analyze the correlation. Motions areaccurately predicted for the spatial constraints represented in theparametric space, and they therefore have few undesirable artifacts,if any. This property alleviates the problem of spatial inconsistencies,such as foot-sliding, that are associated with many existingmethods. Moreover, numerical estimates for the reliability of predictionsenable motions to be adaptively sampled. Since the interpolationkernels are computed with a linear system in real-time,motions can be interactively edited using various spatial controls.