一种可靠的小波去噪质量评价指标

A Reliable Evaluation Indicator of Wavelet De-noising

针对传统评价指标应用于小波去噪质量评价的局限性,提出了一种复合评价指标。将各备选参数的去噪信号的均方根误差与平滑度进行归一化操作,利用变异系数定权的方法将两种指标线性组合,所得到的新指标即为复合评价指标。该指标值越小,表明其去噪效果越好,所选参数越优。理论分析证明该指标具有明确的几何和物理意义,确定了信号细节信息和逼近信息的最佳比例;多组数据分析表明该方法所确定的最佳去噪信号效果较好,简单快速且准确率高。因此,该方法是一种可靠的小波去噪质量评价指标,可以解决小波分析中诸如分解层次、小波基函数等最优参数选择的问题,服务于工程实践。

Aiming at the limitations of existing quality assessment, a composite evaluation is RMSE, smoothness of every de-noising signals; traditional evaluations applied to wavelet de-noising proposed. Firstly, this method will normalize the secondly, the two indicators are linear combinations by the use of coefficient of variation given right~ lastly, the indicators obtained is composite evalua- tion. The index value is smaller, then the de-noising effect is more obvious and the selected parame- ters are better. It has been proved that the indicators have clear geometric and physical meanings through theoretical analysis, besides the best ratio of details and approximations has been determined~ on the other hand, it shows that the determined best de-noising signals have better effects, simple, fast and has high accuracy through multiple sets of data analysis. Therefore, it is a reliable wavelet de-noising quality evaluation and can solve the problems of optimal parameter selection such as decomposition levels and wavelet basis functions in wavelet analysis, which will be serviced in engineering areas.