Methods for estimation of the number of sources in blind source separation
Zhiguo Fan · Journal of Hefei University of Technology · 2008
The number of sources in blind source separation is studied.In the noise-free blind source separation,the method of estimating the source number by null space!of the observed data matrix is proved to be equivalent to the method of computing the rank of the observed data matrix.In the blind source separation when there is noise present in the data,the methods of estimating the number of sources are mainly as follows: eigenvalue decomposition,the Akaike information criterion(AIC),the minimum description length(MDL) principle and the Minka Bayesian criterion.The applicable scope of these estimation methods and main parameters affecting estimation are obtained by theoretic analysis and experiment,which provide a reference for obtaining the number of sources exactly.