Adaptive wavelet denoising method based on sample entropy

It is a key problem of properly selecting and determining the threshold value for wavelet denoising.In this paper,a new method is proposed to adaptively determine the threshold for wavelet analysis.The method uses the character of sample entropy and combines it with wavelet denoising method.Using the method,the sample entropy values are computed for noise time series with different threshold values.The relationship between the sample entropy values and the corresponding threshold values can thus be obtained.As the result,the threshold value is determined at the point with maximum entropy.The method is tested using the synthetic data and annual runoff series from Yingluoxia on the Heihe River basin and Huayuankou on the Yellow River basin.Results show that the method is able to separate the signal and noise from hydrological time series with a good denoising performance.The resulting signal is able to retain the characteristics of hydrological time series and evaluation indexes.The method also offers a new approach to determine the threshold value for wavelet denoising.