The blind separation problem where the sources are not independent, but have variance-dependencies is discussed. Hyvärinen and Hurri proposed an algorithm which requires no assumption on distributions of sources and no parametric model of dependencies between components. In this paper, we extend the semiparametric statistical approach of Amari and Cardoso under variance-dependencies and study estimating functions for blind separation of such dependent sources. In particular, we show that many of ICA algorithms are applicable to the variance-dependent model as well. Our theoretical consequences were confirmed by artificial and realistic examples.
|Number of pages||8|
|Journal||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|Publication status||Published - 2004|
ASJC Scopus subject areas
- Theoretical Computer Science
- Computer Science(all)