To　solve　the　problem　of　insufficient　autonomous　operation　capability　of　intelligent　service　robots,　the　human-in-the-loop　strategy　is　a　useful　mean　to　apply　robotic　technology　for　practical　application.　This　paper　introduces　an　attitude　algorithm　of　a　wearable　human　motion　capture　system　we　designed　to　control　humanoid　robot　to　perform　human-like　action.　First,　the　design　of　this　human　capture　system　consisting　of　multi　inertial　navigation　parts　was　described.　Then,　an　attitude　solution　algorithm　for　this　system　was　put　forward,　which　inputs　human　postures　and　output　accurate　human　joint　data　by　the　quaternion　method　and　the　unscented　Kalman　filter　(UKF).　Experiments　for　attitude　data　analyses　and　telecontrol　of　humanoid　robot　were　performed.　The　results　show　that　the　system　and　its　attitude　algorithm　can　realize　human　joint　angle　calculation　with　high　precision　during　motion,　and　can　be　applied　to　the　teleoperation　of　humanoid　robot.