Aiming　at　the　non-stationary　and　non-linear　characteristics　of　motor　bearing　vibration　signal,　a　fault　analysis　and　identification　method　of　motor　bearing　based　on　extreme-point　symmetric　mode　decomposition　(ESMD)　and　support　vector　machine　(SVM)　is　proposed.　ESMD　uses　the　optimal　adaptive　global　mean　to　determine　the　optimal　number　of　modal　decomposition.　The　best　IMF　component　of　fault　feature　can　be　obtained　by　decomposing　the　vibration　signal　of　motor　bearing　by　ESMD　algorithm.　The　energy　of　IMF　components　which　contain　the　main　fault　features　are　extracted　and　normalized.　The　feature　vectors　are　imported　into　SVM　classification,　and　then　the　fault　types　are　identified　based　on　SVM　classifier.　It　is　verified　by　simulation　experiment　and　database　of　Case　Western　Reserve　University.　Experimental　results　show　that　compared　with　empirical　mode　decomposition　(EMD),　this　method　can　not　only　reduce　the　useless　IMF　components,　but　also　effectively　improve　the　signal　decomposition　accuracy　and　suppress　the　mode　aliasing　phenomenon　in　the　process　of　EMD　decomposition.　Compared　with　the　EMD-SVM　method,　the　ESMD-SVM　based　method　has　higher　accuracy.　This　method　can　provide　a　new　idea　for　motor　bearing　fault　analysis　and　identification.　©　2021,　Politechnica　University　of　Bucharest.　All　rights　reserved.