Signal modulus design for blind source separation via algebraic known modulus algorithm: A perturbation perspective

Abstract

This brief considers blind signal source separation via algebraic known modulus algorithm. It is shown that through proper sign at modulus design the estimation accuracy of the beamforming vector, as well as the performance of signal separation, can be improved. Specifically, based on a matrix perturbation analysis we propose a criterion, in the form of minimizing the maximal singular value of the modulus code matrix, for enhancing robustness of the beamforming vector against measurement noise. A closed-form solution is then derived and its performance is tested through numerical simulation.