论文信息 - Nonnegative Matrix Factorization for Independent Component Analysis

Nonnegative Matrix Factorization for Independent Component Analysis

In this paper, we develop a new algorithm with improved efficiency for nonnegative independent component analysis. This algorithm utilizes Kullback-Leibler divergence to generate nonnegative matrix factorization of the observation vectors. During the factorization, by pre-whitening the observations and orthonormalizing the mixing matrix, the independent components of sources are obtained. In the simulation, we successfully apply the developed algorithm to blind source separation of three images where sources are statistically independent.

Shangming Yang | Shangming Yang | Zhang Yi | Zhang Yi

[1] H. Sebastian Seung,et al. Learning the parts of objects by non-negative matrix factorization , 1999, Nature.

[2] Pierre Comon,et al. Decomposition of quantics in sums of powers of linear forms , 1996, Signal Process..

[3] Patrik O. Hoyer,et al. Non-negative Matrix Factorization with Sparseness Constraints , 2004, J. Mach. Learn. Res..

[4] Terrence J. Sejnowski,et al. An Information-Maximization Approach to Blind Separation and Blind Deconvolution , 1995, Neural Computation.

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

[6] Mark D. Plumbley. Conditions for nonnegative independent component analysis , 2002, IEEE Signal Processing Letters.

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

[8] Erkki Oja,et al. Blind Separation of Positive Sources by Globally Convergent Gradient Search , 2004, Neural Computation.

[9] Erkki Oja,et al. A "nonnegative PCA" algorithm for independent component analysis , 2004, IEEE Transactions on Neural Networks.