Flexible Independent Component Analysis Algorithm Based on Generalized Gamma Distribution
Jia Hairong · Journal of Taiyuan University of Technology · 2009
The flexible independent component analysis algorithm based on Laplacian distribution had slow convergence speed.To improve convergence speed of the algorithm,a flexible independent component analysis algorithm based on generalized gamma distribution was proposed.Taking generalized gamma distribution function as a statistical estimation model for speech signals,a nonlinear activation function more suitable for speech separation was obtained.And on the basis of this function,using this nonlinear activation function in natural gradient based ICA algorithm in computer simulations,we got the results which proved the above mentioned properties.