Local binary patterns-based facial expression recognition using supervised laplacianfaces
Wei Wei · Journal of Optoelectronics·laser · 2008
Nowadays,many approaches based on manifold can't reflect accurately the underlying structure of expressional manifold,so their recognition accuracy is not high.Aimed at this problem,this paper proposed a method based on supervised Laplacianfaces(SLAP)and local binary patterns(LBP).LSLAP is short for the proposed method.The advantages of SLAP are preserving locality and considering the class information into account,which make the different sample classes well separated in the subspace of SLAP.LBP have the advantages of discriminative ability,illumination resistance and simple computation while LBP descripe face image in three different levels of locality.By projecting LBP features onto SLAP subspace,LSLAP extracts the final features and then uses the nearest-neighbor classifier.Experiements on JAFFE and Cohn-Kanade facial expression databases demonstrate that SLAP can reflect accurately the expressional manifold structure and LSLAP is superior to Eigenfaces,Fisherfaces,Laplacianfaces and their respective application to LBP,i.e.,Eigenfaces+LBP,Fisherfaces+LBP,Laplacianfaces+LBP,in terms of recognition accuracy.