MetaCartSign in to MyCiteSeer

Include Citations | Advanced Search | Help

Include Citations | Advanced Search | Help

  Identifiability Issues in Noisy ICA

Download:
Download as a PDF
unknown authors
http://www.elec.qmul.ac.uk/people/miked/documents/IdentifiabilityissuesinnoisyICA.pdf
Add To MetaCart

Abstract:

Abstract—We consider the identifiability of the statistical model for noisy independent component analysis showing that while the mixing process is identifiable, the noise covariance is only partially so. This raises questions as to the performance of certain maximum-likelihood algorithms for blind source separation in the presence of noise. Index Terms—Identifiability, independent component analysis (ICA), noise. I.

Citations

1232 Matrix Computations – Golub, Loan - 1989
867 Independent component analysis, a new concept – Comon - 1994
806 The Algebraic Eigenvalue Problem – Wilkinson - 1965
232 signal separation: Statistical principles – Blind - 1998
166 Independent factor analysis – Attias - 1999
24 Independent component analysis in the presence of gaussian noise by maximizing joint likelihood – Hyvarinen - 1998
17 Learning sparse multiscale image representations – Sallee, Olshausen - 2003
10 Identifiability and separability of linear ICA models revisited – Eriksson, Koivunen - 2003
8 Audio source separation – Davies - 2001
5 A technique for higher-order-only blind source separation – Lathauwer, Moor, et al. - 1996
4 Maximum likelihood for blind signal separation and deconvolution of noisy signals using mixture models – Moulines, Cardoso, et al. - 1997
4 Globally convergent blind source separation based on a mulituser kurtosis maximization criterion – Papadias - 2000
1 moments for noisy independent component analysis – “Gaussian - 1999
1 ICA for noisy data using Gaussian moments – “Fast - 1999
View or Download | Add to My Collection | Correct Errors