Abstract:The minimum phase presumption of seismic wavelet offers a premise for us to solve ill-posed deconvolution problem; however,its correctness is doubtful. Most geophysicists have been analysing the problem only by using second-order statistical informations of seismic data,namely autocorrelation function or power spectrum of seismic data. Second-order statistics have no phase informations,so that they will not be used to solve physical problem of non-minimum phase.The introduction of high-order accumulated amount and multispectral conception makes it possible to identify non-minimum phase wavelet by using the phase informations of high-order accumulated amount or multispectra,or their slices (or projections).A method for obtaining non-minimum phase wavelet by taking a mov- ing-ave=aae parameter estimation based on high-order accumulated amount is recommended in this article. This method makes full use of phase information in seismic accumulated high-order amount,so that the neglect of physical essential which results from the only dependence of optimi2ing thecnique is avoided satisfactorily.Processing result of real data shows that the method is obviously superior in the processing of mixed-phase and non-Gaussian signals.