Abstract:

ICA is a novel stat istical method developed recently, which is used to find a representat ion of the Non-Gaussian mult ivariate data. The representation shows that each component of the computed vector is independent stat istically, or as independent as possible. In application, this kind of transformat ion aims to capture the basic structures of the analyzed data, including features abst raction and separat ion of signals. Presents the fundamental theory and fast algorithms, at the same time, implement the Fast ICA and itsupdated version. Compared w ith PCA or K-L t ransformat ion, proposes the concept of Independent component transformat ion ( ICT). On the basis of analyzing the features of seismic signals, does preliminary studies and try to apply ICA on seismic sig nal processing. Research results show the good perspect ive of ICA application to seismic signal processing.

Key words: independent component analysis, fast algorithm, transformation, seismic signal noise elimination, attempt