earticle

영어

Wavelet threshold filter is widely used because it is simple to implement and small computation. Traditional hard and soft threshold functions have their limitations. At the same time, design method of the threshold function is used only for a particular problem and general construction method of threshold function is rarely involved. In this paper, we study the basic requirements of the threshold function and the threshold objective function, and propose a general method for constructing the threshold function. The y=x method applies rotation, shift and other operation on the basic function, and takes as the objective function, constructs threshold function families with different thresholds and approximation rate. Then, choose different threshold functions according to the characteristics of the signal. According to this approach, we take y(x)=a/x as basic function and construct a new threshold function family. The new threshold function is continuous and differentiable at high order, which overcomes the shortcomings of hard threshold function and soft threshold function. Simulation results show that in different noise levels and different wavelet decomposition layers, it has different noise characteristics; signal-to-noise ratio (SNR) gain and minimum squared error (MSE) of this method is better than that of the traditional soft and hard threshold function and improved threshold function and made a better balance between the sensitive degree and smoothness.