Monitoring　of　potential　bearing　faults　in　operation　is　of　critical　importance　to　safe　operation　of　high　speed　trains.　One　of　the　major　challenges　is　how　to　differentiate　relevant　signals　to　operational　conditions　of　bearings　from　noises　emitted　from　the　surrounding　environment.　In　this　work,　we　report　a　procedure　for　analyzing　acoustic　emission　signals　collected　from　rolling　bearings　for　diagnosis　of　bearing　health　conditions　by　examining　their　morphological　pattern　spectrum　(MPS)　through　a　multi-scale　morphology　analysis　procedure.　The　results　show　that　acoustic　emission　signals　resulted　from　a　given　type　of　bearing　faults　share　rather　similar　MPS　curves.　Further　examinations　in　terms　of　sample　entropy　and　Lempel-Ziv　complexity　of　MPS　curves　suggest　that　these　two　parameters　can　be　utilized　to　determine　damage　modes.