Higher order statistics for the seismic wave identification

Seismic prospecting is used to identify the series of contrasts in the acoustic impedance associated with different formations in the subsoil. Generally, the second order statistics or spectral matrix of the recorded data is used. The eigenvectors of this matrix are partitioned into two orthogonal subspaces: the signal subspace and noise subspace. The problem of this method is the correlation between the signals and the use of the analysis bandwidth because the seismic data are wide band signals. These limit its application on the real data. We propose a novel and efficient algorithm for estimating the wave parameters. This method is based on an extension of a narrow band wave processing in the frequency domain to the wide band signals. The focusing method used with the spectral matrix, is extended to the higher order statistics.