signal processing based on mud pulse is one key of technologies of MWD system. In this paper, the characteristics of mud pulse channel and its noise is analyzed, signal filtering model and identification model is established, one signal processing method consists of wavelet transform, correlation filtering, pattern, recognition and clustering is given for mud pulse signal data processing under bad working environment, which avoid to limitations of bad realtime performance and bad influence from the threshold Field test results show that the method can decode down hole measurement data effectively and improve the signal recognition rate and realtime performance.