Nonlinear constrained optimization using Lagrangian approach for blind source separation

The paper deals with the blind source separation problem. We introduce two new adaptive algorithms based on the minimization of constrained contrast functions using a Lagrangian approach. The algorithms "only" require one stage for separation and the approach is general in the sense that it can be used with any contrasts working with normalized vectors. The computer simulation shows good performances in comparison to the EASI algorithm.