Analyzing and Segmenting Finger Gestures in Meaningful Phases

This paper presents a method to segment the motion of a character's fingers in an efficient way, so that each segmented motion will retain its required meaning. To determine the meaning of a gesture, the motion sequences are analyzed, and the correlation between the finger and the wrist motion is determined. This correlation is quite important, especially in cases that require automatic synthesis of finger animation. The proposed meaningful phases of the finger gestures are evaluated against a standard segmentation method. This involves, firstly, computing the cost of transition between two motion segments and, secondly, using a previously proposed distance metric for automatic finger motion retrieval. As the results show, incorrect calculations, as well as the transition cost, are reduced by the proposed motion segmentation technique.