The present study investigated Frontal EEG feature from differentiated emotion states. In the first study, we initially collected more than one thousand film excerpts about the eight emotion categories: joy, amusement, tenderness, disgust, anger, fear, sadness and neutrality. Then three judges rated each segment and chose 111 excerpts for experiment. After large sample subjects viewed and assessed the film clips, we selected 22 film excerpts which could successfully and specially elicit high target emotion. The second study recorded the EEG data when the participants watching the 16 film clips selected from our standard films database. We analyzed the time-frequency features differences on frontal cortex of the best 8 videos. We examined the frontal alpha asymmetry (FAA) and frontal midline theta (Fm theta), and compared three-pair emotions: amusement-tenderness (positive emotion- positive emotion), tenderness-anger (positive emotion-negative emotion), and anger-fear (negative emotion-negative emotion). We found larger relative left frontal activation in tenderness and larger relative right frontal activation in right frontal, which supported the valence hypothesis, instead of approach-withdrawal hypothesis. Fm theta power positively correlated with valence, which high valence film clips induced high Fm theta power. The differences between in positive emotion and positive emotion, positive emotion and negative emotion were delta, theta, alpha and beta, but in different sub cortices of frontal cortex. And the differences between anger and fear were low bands at the whole frontal cortex: delta and theta. We didn’t find the gamma power differences in these emotion paired comparisons. Besides, we found frontal theta asymmetry (FTA), relative high right frontal activation in positive emotion like tenderness, and relative high left frontal activation in negative emotion like fear and fear. There was no FTA effect in neutrality. Partial correlation results indicated that, compared with valence, the FTA index significantly correlated with liking, larger index in emotions with higher liking.