Blind seismic wavefield separation using frequency singular value decomposition

This paper presents a new blind statistical approach based on frequency singular value decomposition to enhance the SNR of the full multi-component seismic wavefield as well as separating the seismic primary waves. A model of wideband polarized seismic wavefield that are received by linear array of three component sensors is used as framework for implementing the proposed algorithm. This algorithm explicitly exploits the eigen-structure of reduced dimensional spectral covariance matrix. The blind separation of first primary wave is achieved by projecting the first eigenvector that has the highest eigenvalue of this covariance matrix on the long data vector that contains information on all frequencies and all components interactions of the multicomponent seismic wave-field. In addition, the experimental results have shown that the proposed algorithm outperforms the conventional separation technique in terms of accuracy and complexity.