Abstract:We propose a combinational adaptive noise reduction method combining empirical wavelet transform (EWT) and non-local mean (NLM) filtering with sample entropy (SE) optimization . This method uses SE to determine the low-frequency effective signal limit of all empirical modal components, superimposes the remaining medium and high-frequency components, and performs NLM filtering. Finally, the filtered signal and the effective signal are reconstructed as the final noise reduction signal to filter high-frequency noise. Using simulated data and measured data for experimental research, the results show that the optimized EWT-NLM method is overall better than the EMD and EWT methods. The RMSE decreases by 13.41%/10.63%(measured data/simulated data), 7.13%/5.78%, and the signal-to-noise ratio increases by 22.03%/22.54%, 9.72%/7.42%.