This　paper　presents　a　semantic　segmentation　algorithm　under　the　guidance　of　skeleton　of　3D　humanoid　model.　The　algorithm　is　in　two　phase:　(1)　extract　the　skeleton　of　humanoid　model;　(2)　identify　the　feature　points　of　the　extracted　skeleton　and　segment　the　model　according　the　key　point　employ　the　short-cut　rule　of　cross　section.　Due　to　the　complexity　of　calculating　the　general　potential　field,　general　potential　field　skeleton　extraction　algorithm　is　time　consuming.　In　skeleton　extraction　stage　a　curve-skeleton　extraction　algorithm　combining　distance　transform　and　potential　field　is　used　for　less　complexity　computation.　The　algorithm　is　simply　and　easily　operated　and　can　segment　the　humanoid　3d　model　semantically.　©　2011　AICIT.