In this study, a wavelet packet denoising algorithm based on complementary ensemble empirical mode decomposition(CEEMD) was proposed to solve the problem that traditional denoising methods of heart sounds generally eliminate high frequency information and thereby cause signal distortion. Firstly, heart sounds were decomposed into different intrinsic mode functions (IMFs) with CEEMD, and autocorrelation function was utilized to define the range of mode components objectively. Then, the useful information was extracted from noise dominant mode components and mixing mode components using wavelet packet transform and used for reconstruction of the denoised signal with the remaining IMFs.The results showed that the proposed method could not only improve the denoising indexs (SNR and RMSE) significantly afer denoising, but also effectively preserve high-frequency useful information of the signal, have better denoising performance under different noise levels compared with traditional algorithms and better robustness.