Abstract: Aiming at suppressing the strong stochastic noise in seismic prospecting data, support vector regression(SVR) is introduced. A new permitted support vector kernel function-Ricker wavelet kernel function is proposed and demonstrated. Based on the elementary idea of kernel mapping and the principle of structural risk minimization, SVR transforms the regression problem into a quadratic programming problem. The results of simulation experiments for single channel data or arbitrary channel of multi-channel data show the clearer event, the better wave shape and the higher SNR compared with the conventional convolution filter and common SVR based on RBF. So it is possible that SVR based on Ricker wavelet kernel function is applied to suppressing noise in seismic prospecting data.

Key words: support vector regression, Ricker wavelet kernel function, convolution filter, seismic prospecting event