| |
Abstract:
The problem of local independent component analysis is
formulated and solved in this paper. In contrast to iterative
non--local parametric density estimation algorithms for source
separation, here only local information is used and an analytic
solution for source separation is obtained. The local nature of
this approach allows for separation of general smooth invertible
mixing transformations, while the analytic nature allows for a
proper treatment of source separation in the presence of
uncertainty. The mathematical framework presented extends
independent component analysis to a general factorization of
multivariate functions. Numerical examples and a tie to
decorrelation are presented.
|
