Simultaneous Source Separation Using Adaptive Robust Linear Algebra
C. Beasley, I. Moore, R. Fletcher and C. Castellanos
Event name: 78th EAGE Conference and Exhibition 2016
Session: Simultaneous Sources
Publication date: 31 May 2016
Info: Extended abstract, PDF ( 1.15Mb )
Price: € 20
The benefits of simultaneous source acquisition are well known and while we envision processing without separation of sources, today the main challenge is in the quality of the separation. A variety of methods have been developed for separation, however, when source intervals are large and interference is heavy, these methods are challenged and may experience leakage from one source into another in the separation process. Such leakage manifests as noise and can be detrimental to the data. We introduce a new concept in linear algebra that improves separation, particularly for such difficult situations. We introduce the notion of a robust inner product (RIP) which is insensitive to interference encountered in the source separation. This RIP is then used to define robust linear algebra operations. Specifically, inversion approaches to the source separation problem generally solve matrix equations as a sparse minimization problem. We apply the robust approach to the solution in the form of a conjugate gradient iteration, but modified with robust operators. Finally we illustrate the effectiveness of our approach on field data. While adaptive robust linear algebra is very effective for this problem, it is not specific to simultaneous sources. We expect to find a wide variety of applications.