Capturing Natural Hand Articulation

Abstract

Vision-based m, otion captu.ring of hand articulation is-a ch, allengin, g task, since th, e hand presents a m, o-tion of high, degrees of freedom.. Model-based approach, es could he taken to approach this problem. by searching in, 0. h, igh dim.en, sional hand state, sppace, a, nd m, atch,-in, g projection, s of a hand m, odel and im, age ohsemia-tion.s. fIouieaer, it is h, a, qhhj ineficien, t due to th, e curse of dim.ension, ality. Fortumately, nmturd h, and nrticu.la-tion is highly constrained, iihich, lar9ely redu.ces the di-m, ension, nlity of hand.state space. Th'is paper presen, ts n, m.odel-hased m, ethod to captu.re h, an.d articu.lation, by leamin, g h, o, n.d nmturd constmin, ts. Our study sh, ouis th, at natu, ro.l h, an, d articu.lo.tion lies in a lower dimfen-sion, nl config, u.ration, s space charoxterized by a maon, of linenr m, an, ifolds spnn, n, ed by 0. set of basis con~jigu.ra-tion, s. By inte, qratin, g h, o, n, d m, otion, constrain, ts, an, ef-ficien, t articirlnted m.otion,-co, pturin,,q algorithm, is pro-posed hnsed on seque Mmte Co.rlo tech, n, iques. 0u.r en:pcrim.en, ts shoai th, at th, is algorithm. is rohwst and ac-cumte for tro, ckdn,,q nmtural h, nn, d m, ovem, ents. Th, is al-,qorith, m, is ~O,S!J to extend to der articdn.ted m.otion co.ptii, rin, g tasks. 1