Blind source separation based on optimally selected estimating functions

The problem of blind source separation based on the independent component analysis is considered. A algorithm is proposed, which uses the estimating function with faster convergence in the initial stage and the asymptotically optimal one with better steady-state performance in the tracking stage. The proposed algorithm is simple enough to overcome the contradiction inherent in existing approaches between convergence speed and steady-state performance. The simulation results are presented to demonstrate its effectiveness.