Surface　electromyography(sEMG)　signals　mainly　contain　noise　from　around　environment.　The　noise　leads　the　problems　of　poor　recognition　rate　of　lower　limb.　In　this　paper,　a　novel　method　was　proposed　to　processing　sEMG　signals,　named　empirical　mode　filtering　and　self-enhancement　algorithm　with　classical　wavelet　(ESECW),　which　denoises　interface　noise　of　originnl　signals　and　enhanced　the　recognition　rate　of　lower　limb　motions.　The　ESECW　algorithm　consist　of　two　parts:　the　raw　signal　was　first　processed　by　the　empirical　mode　decomposition　(EMD).　In　the　second　part,　the　background　noise　of　the　original　signal　is　firstly　reduced　with　the　band-pass　filter,　and　then　processed　by　four　level　wavelet　decomposition　to　obtain　four　layer　of　high-frequency　signal　components　which　use　to　calculating　average　energy　of　signal.　Finally,　the　above　two　processed　signals　are　multiplied　together　to　obtain　the　mixed　signal.　Thus,　the　active　segment　of　raw　signals　is　derived.　then,　the　extracted　feature　set　is　input　to　the　SVM　for　lowe　limb　motion　pattern　recognition.　The　actual　experimental　analysis　indicate　that　ESECW　method　has　good　adaptability　to　various　motion　patterns　and　does　not　to　set　a　series　thresholds.　©　2021　IEEE.