The　wavelet　transform　de-noising　method　based　on　the　threshold　shrinkage　is　widely　used,　however,　the　classical　threshold　shrinkage　method　maybe　exists　constant　deviation　or　appears　additional　concussion　after　reconfiguration.　If　the　threshold　value　is　chosen　too　large,　part　of　useful　signal　points　will　be　eliminated;　if　chosen　too　small,　part　of　the　noise　will　be　kept.　In　addition,　the　＇first　generation　wavelet＇　de-noising　method　only　can　choose　one　type　of　wavelet　base,　and　can＇t　be　changed　during　the　de-noising　process.　Due　to　the　different　wavelet　has　its　own　characteristics　and　applicable　signal,　It　can　achieve　local　optimum　then　overall　optimization　that　choosing　wavelet　in　a　small　wave　focus　according　to　the　partial　feature　of　the　signal,　namely　using　the　multiwavelets　to　comprehensively　decompose　signal.This　paper　puts　forward　a　new　threshold　function,　then　improves　the　selection　scheme　of　threshold　value,　achieves　that　it　can　select　dynamically　according　to　the　wavelet　decomposition　levels;　Finally,　on　the　basis　of　the　lifting　theory,　this　paper　puts　forward　an　adaptive　method,　which　can　select　wavelet　dynamically　according　to　the　partial　feature　of　the　signal,　comprehensively　making　use　of　the　advantages　of　each　wavelet,　to　achieve　better　de-noising　effect　and　a　certain　degree　of　adaptability.　Experiments　show　the　effectiveness　of　our　optimization.