Wavelet thresholding using higher-order statistics for signal denoising

The paper demonstrates a higher-order statistics (HOS) based method of wavelet thresholding for signal denoising. We calculate the triple correlation coefficients of wavelet-signal correlation for identification of wavelet coefficients uncorrupted by noise. Since the higher than second-order moments of the Gaussian probability function are zero, the Gaussian noise can be eliminated completely. The method is also valid for unknown spectral density noise. The results of computer simulation show the availability and the effectiveness of the proposed wavelet thresholding method.