Wearable　motion-capture　systems　offer　promising　avenues　for　human　lower　limb　rehabilitation.　However,　unstable　data　transmission　and　attitude　estimation　challenge　their　practical　application.　Aiming　at　this　problem,　a　reliable　method　utilizing　wearable　inertial　sensors　for　rehabilitation　applications　is　innovatively　proposed　and　implemented　within　our　designed　wearable　motion-capture　systems　tailored　to　patients　with　impaired　lower　limbs.　A　stable　data　transmission　process　based　on　star-type　Bluetooth　body　sensor　networks　is　designed　by　establishing　a　connection　parameter　setting　method　to　guarantee　reliable　attitude　estimation.　Then,　a　robust　attitude　estimating　method　based　on　an　improved　gradient　descent　method　is　proposed　to　promote　the　anti-interference　capability　of　the　algorithm　by　introducing　trust　coefficients.　Lower　limb　motion-capture　experiments　are　conducted,　and　results　show　that　the　proposed　method　enables　the　system　to　maintain　a　package　loss　rate　of　no　more　than　0.24%　and　has　a　maximum　coefficient　of　variation　(CV)　of　5.9%　during　the　data　transmission　process.　Attitude　estimation　reliability　experiments　reveal　that　the　proposed　algorithm　substantially　enhances　anti-interference　capabilities　while　preserving　estimation　accuracy.　Compared　to　the　state-of-the-art　method,　under　acceleration　shock,　estimation　errors　decrease　by　up　to　39.1%　(roll),　42.9%　(pitch),　and　20.2%　(yaw).　When　exposed　to　external　magnetic　field　interference,　conventional　estimation　algorithms　falter,　whereas　the　proposed　method　maintains　an　average　error　within　2　degrees.　Significance　analysis　underscores　the　method＇s　distinctiveness　at　the　0.05%　significance　level　(p　＜　0.05).　This　study　effectively　bridges　the　gap　between　wearable　inertial　motion-capture　systems　and　their　application　in　clinical　lower　limb　rehabilitation.