Blind Source Extraction with Adaptive Learning Rate Based on a Linear Predictor

An on-line BSE algorithm with an adaptive learning rate is proposed. By indirectly studying one of the existing on-line BSE algorithms based on line predictability, the bound for the optimal learning rate which guarantees the convergence of the algorithm is derived. Based on the analysis results, an on-line algorithm with an adaptive learning rate is presented. Since the learning rates of the existing on-line algorithms based on line predictability are chosen empirically in practice, the adaptive one proposed in this paper is more useful. The simulations verify the obtained results.