Kinect,　as　an　individual　feeling　surrounding　peripherals,　not　only　changes　the　users＇　entertainment　experience,　also　provides　a　new　way　between　users　and　machine　interaction.　Using　easily　and　improving　the　gesture　recognition　accuracy,　the　paper　presents　a　method　that　a　device　sensors　are　used　to　get　the　more　than　20　joint　point　information,　and　a　specific　position　is　distinguished　by　the　Euclidean　distance　and　Angle　of　each　joint　point.　The　result　shows　that　the　recognition　rate　is　higher　using　the　method　to　identify　the　position　and　extension　action　library　is　used　to　meet　the　needs　of　determining　different　positions　at　any　time.　On　this　basis,　the　paper　has　been　implemented　in　various　fields　or　carried　on　the　beneficial　exploration　of　possible　application.　©　(2014)　Trans　Tech　Publications,　Switzerland.