This　article　selects　five　common　classifiers　in　music　emotion　recognition,　namely　SVM　(kernel　function:　RBF),　SVM　(kernel　function:　LINEAR),　kNN　(K　=　3)　and　kNN　(K　=　5),　as　well　as　GMM,　application　active　learning　performs　music　emotional　classification.　The　experiment　uses　365　music　clips　in　the　PMEmo　data　set,　and　after　manual　labeling,　they　are　divided　into　four　groups:　HAPPY,　SAD,　ANGRY　and　RELAX　according　to　the　Thayer　emotion　model　and　the　classification　effect　is　judged　by　comparing　F1-measure.　Experiments　show　that　active　learning　can　reduce　the　need　for　manual　labeling　by　about　78%,　and　it　can　help　to　achieve　better　results　with　fewer　labels.　At　the　same　time,　among　the　three　simple　classifiers,　the　SVM　model　and　active　learning　have　a　better　degree　of　adaptation　in　the　field　of　music　emotion　recognition.