Stable Gesture Verification in Eigen Space

This paper mainly discusses the stable gesture classification criteria, which can identify several human motions from a sequence of images with the eigen space method. We have already proposed the gesture classification technique with the silhouette images and the PCA method, which includes the difficulty in the correspondence between two sets of eigen points, i.e. dictionary and input sequences, which are taken with the different time step. In this paper, a new criteria has been proposed to achieve a stable gesture verification technique with the spline approximation of the input sequence in the eigen space. Experimental results show the validity of this proposed method.