Wavelet-Based Threshold De-Noising Method Improvement And Its Evaluation Study

In the study of the noise filter for the observed time series, the developed semi-soft threshold wavelet de-noising methods has achieved better noise reduction effect in recent years. But the difficulty is how to determine the optimal parameters of the threshold value and adjustment coefficients. In this paper, a new semi-soft threshold function is constructed, the function refining treatment the wavelet coefficients which within the threshold value, the method is much more accord with the complexity and continuity characteristic of the natural signal. And a chaotic search method is introduced, the optimum estimation of the wavelet coefficients can be obtained between the hard and soft threshold functions.1) The general principle of the wavelet bases functions and threshold value selection is established. As the wide range of the wavelet bases, the characteristic different with each other, the selection of the threshold value is not unique. In the de-noising process, different wavelet bases and threshold value are used on the same noisy signal, after a large number of simulation comparisons, a general principle of selecting the wavelet bases function and threshold values is obtained. At the same time, three quantitative evaluation criteria is established, including signal to noise ratio, mean square error and smoothness, which can reflect the effectiveness of the de-noising result from different aspects.2) Adjustable parameters semi-soft threshold de-noise study with chaotic search method. A semi-soft threshold function is proposed, the parameters of the function can be adjusted adaptively with different noisy signal, so that more detailed components of the signal can be extracted. At the same time, in accordance with the ergodicity of the Logistic chaotic equation, the de-noising parameters are brought into the iterative process of the chaotic equation, and the ergodicity range of the chaotic motion is enlarged to the scope of the parameters to be optimized. The optimized adjustable parameters can be obtained by ergodicity search using chaotic equation. Simulation results show that using the improved threshold function can extract more detailed signal information, while the denoised signal has better continuity and smoothness.3) The design and development of the wavelet-based threshold de-noising system. System using Microsoft Visual C + +6.0 and MATLAB7.1 languages is consists of user management, the traditional threshold function de-noising and the improved threshold function de-noising modules. The interface of the system is friendly, which is able to visualize the effect of the de-noised signal. The system has some practical value.