As　the　developing　of　computer　application　technology,　batches　of　new　generation　of　somatosensory　devices　emerge.　In　June　2010　Microsoft　release　of　the　kinect　motion-sensing　camera　is　one　of　the　most　representative　somatosensory　devices.　The　kinect　camera,　compared　with　the　traditional　camera,　the　biggest　advantage　of　is　the　depth　camera.　Many　professionals　were　called　depth　camera　kinect　as　“the　third　eye”.　The　depth　camera　can　take　not　only　a　traditional　2D　image　which　have　xy　two　axes,　but　also　the　depth　image　data　which　can　be　used　to　describe　the　z　axis.　By　using　the　depth　image　data　for　image　recognition,　we　can　identify　a　more　accurate　human　behaviour.　This　paper　mainly　expounds　how　to　use　depth　data,　combined　with　quaternion　rotation　to　identify　and　track　a　more　accurate　human　behavior.　©　2014　WIT　Press.