The heart sound signal noise reduction is the premise of heart sound signal for clinical using. A dual self-adaptive lifting algorithm for heart sound noise reduction was proposed. The algorithm used self-adaptive update and self-adaptive prediction to structure wavelet function, and combined the traditional hard threshold with soft threshold function to construct a improved threshold function for the heart sound signal noise reduction processing. 80 cases of heart sound signals from clinical collection of denoising experiments were done. The result showed that it had good denoising effects, and enhanced the local characteristics of the signal . Compared with the ordinary wavelet denoising, the signal-to-noise ratio can be increased by 46.5%, the root-mean-square error was reduced by 64.0%, and running speed was fast. The dual self-adaptive lifting algorithm can process clinical heart sound signal effectively.