the independents component analysis is statistical analysis approach used to solving the blind source separation problems, such as speech signals and/or images
Though your question is not quite clear, it seems to suggest that ICA is a nonlinear technique. It is not. ICA, as PCA, is linear. The difference is that (1) ICA assumes non-Gaussian distribution of data, (2) ICA does not require the resulting vectors of maximal variance to be perpendicular.