This　paper　presents　a　disturbing　recovery　approach　for　lower　extremity　exoskeletons　with　passive　ankle　joints,　which　particularly　works　when　an　external　disturbance　is　applied　on　the　standing　exoskeleton.　Firstly,　by　analyzing　the　mechanical　structure　of　lower　extremity　exoskeletons　during　its　stance　phase,　the　mathematical　model　is　established　using　state-space　equations.　Then,　on　the　basis　of　LQR　(Linear　Quadratic　Regulator)　theory,　a　novel　neural　network　controller　is　designed　to　obtain　the　optimal　solution　for　the　balance　control　of　model　proposed.　Finally,　experimental　results　demonstrate　that　the　proposed　disturbing　recovery　approach　not　only　has　faster　recovery　speed,　but　also　can　achieve　higher　control　precision　than　the　traditional　LQR.