Blind identification of seismic signals

VANDERBAAN Abstract: Seismic processing methods are often based either on wave-theoretical approaches or 2nd order statistics. The project blind identification of seismic signals focuses on developing new techniques based on higher-order statistics. I will present three of our recently devised methods, namely (i) local Singular Value Decomposition (SVD), (ii) blind deconvolution and (iii) blind PP/PS wavefield separation. Local SVD emphasises the lateral continuity of both pre- and post-stack data and has several advantages over standard methods. Blind deconvolution methods are unique in that they can handle bandlimited non-minimum phase wavelets. The new PP/PS separation scheme is unusual in the sense that it requires no a priori information on P- and S-wave velocities - contrary to conventional wave-theoretical approaches. These techniques are not only of interest for processing controlled-source seismic data but also represent promising new avenues to extract more information from earthquake seismological data. ---------------------------------------------------------------------- Séminaires Gésciences Marines ----------------------------------------------------------------------