Independent Component Analysis Based Denoising Technology

The paper introduces a new technology of signal processing:independent component analysis, including its basic concept, principles, and some representative algorithms, such as FastICA, EASI, Nonlinear PCA, and natural gradient algorithm based maximum likelihood estimation. In a denoising simulation experiment with the mean square error criterion, these algorithms are compared to the classic algorithms of adaptive signal processing, such as LMS and RLS. Results show that in denoising application ICA algorithms are superior to the classic adaptive algorithms. Thus ICA algorithms have large value in denoising application, deservnig further study and promoting.