Blind source separation with an unknown number of sources -a semiparametric statistical approach
Jin Hai-hong · Journal of Xidian University · 2003
The problem of blind source separation (BSS) with an unknown number of sources is considered. Firstly, the semiparametric statistical approach is introduced into the BSS, and an estimating function for the semiparametric statistical approach in BSS is proposed, from which a learning rule is obtained. Each separating point is the equilibrium of the learning rule. Finally, the computer simulation shows that the learning rule obtained from the semiparametric statistical approach has better performance in convergence than the one based on the natural gradient.