Abstract:

A novel type of threshold functions with a variable parameter is proposed based on the wavelet shrinkage architecture presented by Donoho  D L and Johnstone I M, and the effect of different parameter on signal shrinkage is also analyzed. This new type of threshold functions has many advantages over the soft and hard threshold functions. It is continuous and simple in expression, and has any order derivative which makes some kinds of mathematical processing very simple and convenient. It also overcomes the shortcomings of the soft threshold method where there is an invariable dispersion between the estimated wavelet coefficients and the decomposed wavelet coefficients. Further more, the new type of threshold functions is more flexible than the softthreshold and hardthreshold functions and integrates their excellence features. All these advantages make it possible to construct an adaptive signal denoising algorithm. Comparative experiment results show that the wavelet shrinkage method adopting the new threshold function is more capable of eliminating the noise from the signal, and results in less MSE and higher SNR than the traditional method adopting hard and soft threshold function.

Key words: wavelet shrinkage;threshold function;MSE;SNR;PSNR