A multi-objective approach for blind source extraction
Guilherme Dean Pelegrina, Leonardo Tomazeli Duarte · 2016
Classically, the problems of blind source separation and extraction are tackled by optimizing a single separation criterion which is associated with a given property of the sources of interest. Motivated by the fact that, very often in practice, there is more than one property to be exploited, the present work introduces a novel separation framework that relies on multi-objective optimization. In order to demonstrate the viability of the proposed framework, we test it in a situation in which both the sparsity and temporality of the source of interest are exploited. Numerical experiments suggest that the worst case Pareto-optimal solution has similar performance compared to the cases where each property is exploited separately via a single-objective formulation.