A novel segmental method for 3D motion data

This paper proposes a novel method that can be used to segment distinct movements from continuous long motion sequences efficiently. Original motion sequences lie on a high-dimensional manifold which is highly contorted, so it is difficult to cluster the similar poses together to form distinct movements. Here we use a dimensionality reduction technique (PCA) to map original motion sequences into low-dimensional subspace, and then we can segment different movements apart in this subspace. We test our method on twenty motion sequences and compare the performance with that of other methods..