Robust　facial　expression　recognition　under　facial　occlusion　condition　is　the　main　research　orientation,　which　has　important　research　significance.　Many　problems　are　caused　by　facial　occlusion,　not　only　missing　facial　expression　information,　but　also　bringing　outliers　or　lots　of　noise.　Aiming　at　the　point,　firstly,　the　face　to　be　recognized　is　reconstructed　using　robust　principal　component　analysis　(RPCA);　secondly,　Eigenfaces　and　Fisherfaces　are　used　to　extract　facial　expression　features　respectively;　finally,　nearest　neighbor　method　and　support　vector　machine　are　used　as　classifiers.　Facial　expression　recognition　experiments　are　implemented　in　different　occlusion　conditions　on　Japanese　female　facial　expression　database　(JAFFE).　On　the　condition　of　big　occlusion　and　small　sample,　RPCA　algorithms　gained　better　recognition　results　than　many　other　methods,　showing　that　this　method　based　on　RPCA　is　robust　to　kinds　of　facial　occlusions.　©　2011　Springer-Verlag.