In　this　paper,　we　extract　face　animation　parameters　from　a　specific　person　with　different　expressions　by　the　method　based　on　3D　scan　data.　By　using　the　Independent　Component　Analysis　for　captured　FAPs　treated　as　training　set,　a　vector　space　representation　is　formed　and　this　space　is　called　general　face　animation　mode　because　it　is　dependent　of　specific　people.　Finally　we　propose　this　space　for　creating　facial　expressions　for　any　general　people.　The　experimental　results　show　such　a　representation　affects　correlation　and　independence　of　every　part　of　a　face　so　that　the　realism　of　face　animation　with　several　expressions　is　improved.