Study on the noise source identification based on a novel variable step-size algorithm of blind sources separation

Some preknowledge of sources input signals or transmission paths were required in advance for traditional noise source identification. In this paper, a novel variable step-size algorithm of blind source separation (BSS) is proposed to identify noise source, which doesn't need any preknowledge but some statistical assumptions about sources. Most BSS algorithms have been issued on fixed step-size, relatively little work has been focused on variable step-size. The output feedback of step-size update in the proposed algorithm is derived from the analysis of adaptive blind sources separation and adaptive filter. With the natural gradient algorithm based minimum mutual information, the iteration formula of separate matrix is also obtained. The availability of this algorithm is confirmed by simulations. Namely, the convergence rate of this algorithm is rapid, while ensuring low steady-state error.