Noisy blind source separation based on adaptive noise removal
Hui Nee Tang, Shu Wang · 2012
A novel natural gradient algorithm with two-step pre-processing is proposed to solve the noise blind source separation problem. That is bias removal techniques followed by a de-noise processing based on least square. Then obtain an algorithm jointly estimating mixing-matrix and decreasing noise. In the condition of that sources are mutually independent and added-noise is independent with any source, computer simulations verify the algorithm is effective and perform better than traditional algorithm.