Abstract:

The approach of estimating the number of signals based on information theoretic criteria has good performance
in the assumption of white noise, but it always leads to false estimation of the coherent sources in colored noise. An approach combining the combined information theoretic criteria and eigenvalue correction, is presented to determine number of signals. The method uses maximum likelihood (ML) and information theoretic criteria to estimate coherent signals alternately, then eliminate the inequality of the eigenvalues caused by colored noise by correcting the noise eigenvalues. The computer simulation results prove the effective performance of the method.