论文信息 - Some experiments on independent component analysis of non-Gaussian processes

Some experiments on independent component analysis of non-Gaussian processes

This paper reports on numerical experiments on the 'independent component analysis' (ICA) of some non-Gaussian stochastic processes. It is found that the orthonormal basis discovered by ICA are strikingly close to the wavelet basis.

[1] Yves Meyer,et al. Wavelets and Applications , 1992 .

[2] J. B. Buckheit,et al. Time-frequency tilings which best expose the non-Gaussian behavior of a stochastic process , 1996, Proceedings of Third International Symposium on Time-Frequency and Time-Scale Analysis (TFTS-96).

[3] Terrence J. Sejnowski,et al. Edges are the Independent Components of Natural Scenes , 1996, NIPS.

[4] Pierre Comon. Independent component analysis - a new concept? signal processing , 1994 .

[5] David J. Field,et al. Emergence of simple-cell receptive field properties by learning a sparse code for natural images , 1996, Nature.

[6] R. Coifman,et al. Entropy-based Algorithms for Best Basis Selection from Which We Can Estimate a L+1 2 2 L . Thus a Signal of N = 2 L Points Can Be Expanded in 2 N Diierent , 1992 .

[7] Jean-Francois Cardoso,et al. Blind signal separation: statistical principles , 1998, Proc. IEEE.

[8] Jean-Franois Cardoso. High-Order Contrasts for Independent Component Analysis , 1999, Neural Computation.

[9] Pierre Comon,et al. Independent component analysis, A new concept? , 1994, Signal Process..