A new signal de-noising algorithm combining improved thresholding and patternsearch algorithm
Xiaojing Chen, Di Wu, Yong He, Liu Shou · 2008
A new de-noising algorithm combining improved thresholding and patternsearch (PS) algorithm was put forward. The improved thresholding method based on Donoho’s method. The traditional wavelet thresholding method includes two kinds: hard-thresholding and soft-thresholding. The hard-thresholding methods may lead to oscillation of the reconstructed signal, and the soft-thresholding methods may cause constant deviations between the estimated wavelet and original wavelet coefficients. The improved threshhoding method can overcome these defects, which was better in keeping trade-off between smoothness and remaining edge of the original signal. A coefficient β which is flexible was set up in the improved thresholding method, how to find the appropriate β is important for the improved thresholding de-noising, furthermore, the improved thresholding methods were combined with some parameters such as wavelet function, decomposition scales etc., the effectiveness of signal de-noising is quite different. Now, most of researchers are usually selected semi-empirically or empirically these parameters, which cannot ensure that the de-noising performance is optimal in some sense. In order to solve these problems, the pattersearch function of Matlab can be adopted to guide the selection of these parameters. The effectiveness of the new method is validated by the results of the simulated experiment.