Chapter 7
Second-order methods based on color
7.1 Introduction
The previous chapters focused on the basic, classical ICA paradigm, in which one of the underlying assumptions is that each of the independent sources can be modeled as a sequence of independent, identically distributed (iid) random variables (possibly, but not necessarily, with a different distribution for each source). As such, the only key for separation in a truly blind scenario is the non-Gaussianity of the sources. Thus, as shown in previous chapters, second-order statistics (SOS) alone cannot yield consistent separation in such models, simply because they are unable to “capture” the non-Gaussianity. They may be used for attaining spatial decorrelation (a.k.a. spatial ...
Get Handbook of Blind Source Separation now with the O’Reilly learning platform.
O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.
