Realtime estimation ofhumanhandposture

Theauthors propose asystem forsearching the shapeofhumanhandsandfingers inrealtimeandwithhigh accuracy, without using anyspecial peripheral equipment such asrangesensor, PC cluster, etc., bya methodofretrieving similar imagequickly withhigh accuracy fromalarge volume ofimagedatabase containing thecomplicated shapes andself- occlusions. Indesigning thesystem, weconstructed adatabase inawaytobeadaptable eventodifferences amongindividuals, andsearched CG images ofhandsimilar tounknownhand image, through extraction ofcharacteristics using high-order local auto- correlational patterns, reduction oftheamountof characteristics centering onprincipal component analysis, and prior rearrange- mentofdatacorresponding totheamountof characteristics. As a result ofexperiments, our system performed high-accuracy estimation ofhumanhandshape wheremeanerror was7degrees infinger joint angles, withthe processing speed of30fpsorover.