We investigate the information processing of a linear mixture of independen
t sources of different magnitudes. In particular we consider the case where
a number m of the sources can be considered as "strong" as compared to the
other ones, the "weak" sources. We find that it is preferable to perform b
lind source separation in the space spanned by the strong sources, and that
this can be easily done by first projecting the signal onto the m largest
principal components. We illustrate the analytical results with numerical s
imulations. (C) 2000 Elsevier Science Ltd. All rights reserved.