Multi-operator based signal separation approach

Abstract

The null space pursuit (NSP) algorithm is an operator-based signal separation approach which separates a signal into a set of additive subcomponents using adaptively estimated operators. In this paper, we introduce a multi-operator based strategy and propose multi-operator based null space pursuit (MONSP) algorithm for signal separation. On the one hand, the proposed approach is a generalized formulation of the original work in some sense. On the other hand, compared with the NSP, the proposed approach has its intrinsic advantages: 1), the proposed approach avoids the mode-mixing problems which is often meet in NSP algorithm; 2), the multi-operator based formulation could theoretically reach the ideal separation resolution for the noise-free signal separation; 3), the MONSP algorithm does not need to initialize parameters in each iteration. We perform several experiments on synthetic and real-life signal separation by MONSP and compare some results with the state-of-the-art methods to demonstrate the efficiency and robustness of the proposed method.