Blind signal separation based on new nolinear function

The independent component analysis (ICA) is one of the most general methods for solving the blind signal separation. It gains lots of applications in communication, speech and medical science. When more sensors are used or the number of sources changes dynamically, natural gradient separation algorithm (NGSA) can solve the problem in a certain limit and the option of nonlinear function affects the convergence and robustness of separating algorithm. In this study, we propose a new nonlinear function applying to NGSA. By setting appropriate step size, the new function can improve the performance of separation algorithm. Simulation results confirm the effectiveness.