Abstract: In a signal，singularities and irregular structures usually take some important information，and wavelet transform is an efficient tool for detecting singularities and edges.In this paper ，according to the range of the cone of influence of singularities and the properties of maxima propagation in wavelet transform domain，a new denoising algorithm is put forward to denoise and preserve the original character of a signal，to seek the wavelet coefficients which are in the cone of influence of each singularity and dispose such wavelet coefficients.The result of experiment shows that the new algorithm implements easily and has better denoising performance.