At　the　present　the　facial　expression　recognition　is　susceptible　to　gender,　nationality　cultural　factors.　In　order　to　get　the　robustness　of　face　recognition　method,　a　method　based　on　the　semantic　and　case-based　reasoning　is　presented.　Firstly,　this　method　automatically　extracted　face　low-level　feature　points　with　DVF,　and　extracted　the　wrinkles　with　LBP.　Then　based　on　the　semantic　AHP　method　layered　facial　expression,　parameterize　expression　semantic　vector.　Three　confidence　estimators　based　k-NN　proposed　to　calculate　the　weights　of　the　semantic　vector.　Finally　design　facial　expression　classifier　based　on　CBR.　The　proposed　method　recognized　facial　expression　in　the　image　high-level　semantic,　is　robust　for　gender,　nationality　culture.　The　experiments　are　conducted　on　Cohn-Kande　database,　the　whole　recognition　rate　of　95.07%　is　achieved.　Theoretical　analysis　and　experimental　results　both　show　that,　the　proposed　method　has　better　recognition　effect　than　others.　©　2010　Binary　Information　Press.