Non-linear Blind Source Separation Using Constrained Genetic Algorithm
In this paper, a novel adaptive algorithm based on constrained genetic algorithm (GA) is presented for solving non-linear blind source separation (BSS), which can both get out of the trap of local minima and restrict the stochastic decision of GA. The approach utilizes odd polynomials to approximate the inverse of non-linear mixing functions and encodes the separating matrix and the coefficients of the polynomials simultaneously. A novel objective function based on mutual information is used with the constraints to the separating matrix and the coefficients of the polynomials respectively. The experimental results demonstrate the feasibility, robustness and parallel superiority of the proposed method.
KeywordsGenetic Algorithm Mutual Information Blind Source Separation Blind Signal Maximum Iteration Number
Unable to display preview. Download preview PDF.
- 2.Gao, Y., Xie, S. L.: An Algorithm for Nonlinear Blind Source Separation Based on Signal Sparse Property and Kernel Function. Computer Engineering and Applications, 22 (2005) 33–35Google Scholar
- 3.Anthony, J. B., Terrence, J. S.: An Information-maximization Approach to Blind Separation and Blind Deconvolution. Neural Computation, 7 (1995) 1129–1159Google Scholar
- 5.Xie, S. L., He, Z. S., Gao, Y.: Adaptive Theory of Signal Processing. 1st ed. Chinese Science Press, Beijing (2006) 136–155Google Scholar
- 7.Zhang, J. L., He, Z. S., Xie, S. L.: Sequential Blind Signal Extraction in Order Based on Genetic Algorithm. Acta Electronica Sinica, 32 (2004) 616–619Google Scholar
- 8.Liu, H. L., Xie S. L.: Nonlinear Blind Separation Algorithm Based on Multiobjective Evolutionary Algorithm. Systems Engineering and Electronics, 27 (2005) 1576–1579Google Scholar