This　paper　proposes　an　external　force　detection　method　for　humanoid　robot　arm　without　using　joint　torque　sensors,　which　can　detect　the　external　force　of　the　joint　space　in　real　time　during　the　operation　of　the　robot.　We　first　analyzed　the　structure　of　the　humanoid　robot　arm　we　designed,　and　then　established　the　external　force　detection　model　of　the　robot　arm　based　on　robot　dynamics　and　motor　dynamics.　Subsequently,　analyses　were　conducted　on　the　error　of　the　detection　model　and　the　dynamic　model　error　of　the　robot　arm　is　compensated　by　using　the　artificial　neural　network　method　to　obtain　more　accurate　external　force　value　for　the　robot　arm.　In　experiment,　the　accuracy　test　and　the　collision　test　were　performed　on　the　detected　extern　forces　of　the　robot　arm.　The　results　show　that　the　method　can　effectively　improve　the　detection　accuracy　of　the　robot　arm,　and　the　robot　arm　can　realize　the　real-time　collision　detection　during　its　static　and　running　states,　which　can　ensure　the　safe　operation　of　the　robot.