New blind separation algorithm using cumulant measure
Bin-Chul Ihm, Kwang-Seop Eom, Dong-Jo Park · 2002
We give a new blind source separation algorithm minimizing a new cost function which is the index of independence of separator outputs. We take the cost function using an infinitely differentiable nonlinear function which has the compressed form of all cross-cumulants. Minimization of the cost function is performed by the parameters adaptation and for fast adaptation, a modified steepest descent method is presented. Computer simulation results provide the validity of the proposed algorithm.