Exterior Penalty Function Method Based ICA Algorithm for Hybrid Sources Using GKNN Estimation
Novel Independent Component analysis(ICA) algorithm for hybrid sources separation based on constrained optimization—exterior penalty function method is proposed. The proposed exterior penalty ICA algorithm is under the framework of constrained ICA(cICA) method to solve the constrained optimization problem by using the exterior penalty function method. In order to choose nonlinear functions as the probability density function(PDF) estimation of the source signals, generalized k-nearest neighbor(GKNN) PDF estimation is proposed which can separate the hybrid mixtures of source signals using only a flexible model and more important it is completely blind to the sources. The proposed EX-cICA algorithm provides the way to wider applications of ICA methods to real world signal processing. Simulations confirm the effectiveness of the proposed algorithm.
KeywordsIndependent Component Analysis Constrain Optimization Problem Independent Component Analysis Natural Gradient Penalty Function Method
Unable to display preview. Download preview PDF.
- 8.Lu, W., Rajapakse, J.C.: Constrained independent component analysis. In: Leen, T., Dietterich, T., Tresp, V. (eds.) Advances in Neural Information Processing Systems, vol. 10, pp. 570–576. MIT Press, Cambridge (2001)Google Scholar
- 11.Silverman, B.W.: Density Estimation for Statistics and Data Analysis. Chapman and Hall, New York (1985)Google Scholar
- 12.Pierani, A., Piazza, F., Solazzi, M., Uncini, A.: Low complexity adaptive nonlinear function for blind signal separation. In: Proc. IEEE Int. Joint Conf. Neural Networks, Italy, vol. 3, pp. 333–338 (2000)Google Scholar