A de-noising method using the improved wavelet threshold function based on noise variance estimation. (15th January 2018)

• Record Type: Journal Article
• Authors: Liu, Hui; Wang, Weida; Xiang, Changle; Han, Lijin; Nie, Haizhao
• Is Part Of: Mechanical systems and signal processing. Volume 99(2017)
• Page Start: 30
• Page End: 46
• Publication Date: 2018-01-15
• Subjects: Wavelet transform -- Gaussian mixture model -- Noise variance -- Electro-mechanical transmission; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2017.05.034
• Languages: English
• ISSNs: 0888-3270

Abstract

Highlights: Two-state Gaussian mixture model to classify high-frequency wavelet coefficients. Multi-scale wavelet transform of noise to estimate the noise variance. An improved wavelet threshold function combining hard and soft threshold functions. Adjusting parameter values of the bivariate threshold function to adjust the values. Abstract: The precise and efficient noise variance estimation is very important for the processing of all kinds of signals while using the wavelet transform to analyze signals and extract signal features. In view of the problem that the accuracy of traditional noise variance estimation is greatly affected by the fluctuation of noise values, this study puts forward the strategy of using the two-state Gaussian mixture model to classify the high-frequency wavelet coefficients in the minimum scale, which takes both the efficiency and accuracy into account. According to the noise variance estimation, a novel improved wavelet threshold function is proposed by combining the advantages of hard and soft threshold functions, and on the basis of the noise variance estimation algorithm and the improved wavelet threshold function, the research puts forth a novel wavelet threshold de-noising method. The method is tested and validated using random signals and bench test data of an electro-mechanical transmission system. The test results indicate that the wavelet threshold de-noising method based on the noise variance estimation shows preferable performance in processing the testing signals of the electro-mechanical transmission system: it can effectively eliminate the interference of transient signals including voltage, current, and oil pressure and maintain the dynamic characteristics of the signals favorably.