基于信噪比评价的阈值优选小波去噪法

Wavelet de-noising method with threshold selection rules based on SNR evaluations

在汽车自动变速器台架试验中,需要对角加速度信号进行消噪滤波。小波阈值去噪是计算量较小、滤波效果较好的消噪手段,但采用不同的小波基匹配不同的阈值规则可能会对去噪效果产生影响。为探寻相对最优组合,该文构建了近似观测信号的仿真信号,然后对该信号做了去噪实验,计算信噪比(SNR)和均方根误差(RMSE),并以此作为评价指标,从而得到相对最优的匹配组合,最后将该组合用于角加速度信号消噪处理过程,取得了较好的滤波效果。小波阈值去噪在处理含噪信号时具有效率较高、稳定性好、不易失真的特点;使用SNR结合RMSE可以对任何消噪结果作客观评判;不同的含噪信号,可能需要用到不同的小波基函数,同时匹配不同的阈值选取规则。

Acceleration signals transmission bench tests need to from automotive automatic be de-noised. Wavelet threshold de-noising requires a relatively small amount of computations and has better filtering. However, different wavelet basic functions and different threshold rules produce different noise signal de noising results. Optimal matching parameters are found for simulated signals that approximate an observed signal through de noising tests using the signal-to-noise ratio （SNR） and root-mean-square error （RMSE） of the de-noised signal as the evaluation index. The results are combined with the angular acceleration signal de-noise processing obtained in a bench test for the filtering. The wavelet transform de-noising is efficient, stable and not easily distorted when dealing with noise signals. The effect of the wavelet de noising noisy signal can be evaluated using SNR combined with RMSE for the objective evaluation. Different signals may need to use different wavelet basis functions with different threshold rules.