An Improved Wavelet De-noising Method for Time Series Analysis

On the basis of discussing some key problems about wavelet de-noising as: choice of reasonable wavelet function, determination of reasonable wavelet coefficients thresholds and choice of suitable threshold processing-means, an improved wavelet de-noising method has been proposed. Then by Monte-Carlo tests, the validity of this method is verified. Analyses results show that compared with traditional methods (FT, SURE and MINMAX), this improved wavelet de-noising method is more accurate and reliable. Furthermore, because of based on information entropy theories to choose the reasonable wavelet coefficients thresholds, the de-noising results by the improved method are the global optimum.