Projection-based classification of surfaces for 3D human mesh sequence retrieval

We analyze human poses and motion by introducing three sequences of easily calculated surface descriptors that are invariant under reparametrizations Euclidean transformations. These obtained associating to each finitely-triangulated two functions on the unit sphere: for vector u we compute weighted area projection onto plane orthogonal length its line spanned u. The L2 norms inner products projections these space spherical harmonics order k provide us with reparametrization invariants surface. use reduces comparison 3D+time representations polygonal curves in R^n. experimental results FAUST CVSSP3D artificial datasets promising. Moreover, a slight modification our method yields good noisy real dataset.