Facial Pose and Expression Analysis Based on Locally Linear Embedding
Chang Liu, Jiliu Zhou, Kun He, Duan Yu-mei · 2008
Variety of facial pose and expression will affect result of face recognition. Traditional methods analyze facial pose and expression through every pixel of image, on the contrary, the paper analyzes pose and expression through geometric relationship among images. The paper uses locally linear embedding (LLE) which computes low dimensional, neighborhood preserving embedding of high dimensional data. The algorithm not only finds the nonlinear structure of data effectively, but also is invariable to translation and revolution. The experimentation shows that the algorithm can reflect slight variety of facial pose and expression, and indicates that low dimensional distribution of various pose and expression is identical to different persons and original image can be reconstructed exactly from neighboring images.