The threshold for wavelet de-noising based on multiple hypothesis test
Du Wen-liao1,2; Liu Cheng-liang1; Li Yan-ming1
1. School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China. 2.School of Mechanical and Electronic Engineering, Zhengzhou Institute of Light Industry, Zhengzhou 450002, China
Abstract:Threshold de-noising method based on the wavelet transform is an effective approach to reduce the white noise in the digit signal. Considering the different characters of signal and white noise in the wavelet domain, a novel algorithm to determine the wavelet threshold is proposed with the multiple hypothesis test. Since the wavelet de-noising transform process can be regarded as a multiple hypothesis process and both the step-up and step-down procedure can control the FDR, the new method FDR step-up-down procedure combining the above-mentioned procedures determines the wavelet threshold. An attractive advantage of this method is it can obtain the desired result by adjusting the FDR level flexibly. The simulated numerical results show that this method works as effective as hearsure method and gives better SNR gains and MSE performance than both the traditional BH FDR and sqtwolog method. The selection of the significance level is also discussed, and point out that the relationship between the significant level and the improvement of the signal to noise ratio is nonlinear, then the tactic of selecting the proper significant level is proposed.
杜文辽; 刘成良; 李彦明 . 基于多假设检验的新型小波滤波算法[J]. , 2011, 30(7): 197-200.
Du Wen-liao; Liu Cheng-liang;Li Yan-ming. The threshold for wavelet de-noising based on multiple hypothesis test. , 2011, 30(7): 197-200.