With　the　advancement　of　data　analysis　and　pattern　recognition　technology,　gesture　recognition　based　on　electromyographic　signals　(EMG)　has　been　widely　used　in　the　field　of　rehabilitation　therapy　and　human-computer　interaction.　Currently,　most　gesture　recognition　models　use　deep　learning　techniques,　but　face　problems　such　as　data　labels　needing　manual　labeling　and　inaccurate　EMG　data　labels,　resulting　in　low　and　unstable　recognition　accuracy.　In　addition,　these　systems　mainly　rely　on　computing　devices,　which　is　not　conducive　to　patients’　rehabilitation　at　home.　Therefore,　this　paper　proposes　a　cloud-based　rehabilitation　system　based　on　automatic　processing　of　sEMG　signals.　The　method　is　based　on　the　acquisition　of　sEMG　data　on　the　surface　of　the　AoYi　bracelet,　applying　an　automatic　annotation　algorithm　to　the　acquired　data,　and　then　uploading　it　to　a　cloud　model　(MotionSenseCNN)　for　training　and　testing,　and　ultimately　realizing　gesture　classification.　Our　network　model　achieves　90%　accuracy　on　the　sparse　channel　EMG　data　collected　in　the　experiment.　The　experimental　results　show　that　the　system　maintains　high　accuracy　in　different　gesture　recognition,　enhances　the　stability　and　flexibility　of　gesture　recognition,　and　has　significant　advantages　over　other　EMG　rehabilitation　systems.　©　The　Author(s),　under　exclusive　license　to　Springer　Nature　Singapore　Pte　Ltd.　2025.