Our　project　focuses　on　the　emotional　face　evoked　EEG　signal　recognition.　Since　EEG　signals　contain　enough　information　to　separate　different　emotional　facial　expressions.　Thus　we　propose　a　new　approach　which　is　based　on　global　field　power　on　EEG　signal　classification.　In　order　to　perform　this　result,　firstly,　we　gather　a　dataset　with　EEG　signals.　This　is　done　by　measuring　EEG　signals　from　people　aged　20-30　that　are　stimulated　by　emotional　facial　expressions　(Happy,　Neutral,　Sad).　Secondly,　the　collected　EEG　signals　are　preprocessed　through　using　noise　reduction　method.　And　then　select　features　by　principal　component　analysis　(PCA)　to　filter　out　redundant　information.　Finally,　using　fisher　classifier　and　a　10-fold　cross　validation　method　for　training　and　testing,　a　good　classification　rate　is　achieved　when　combination　local　max　global　field　power　EEG　signals.　The　rate　is　90.49%.