A novel eigenvalue source enumeration algorithm based on eigenvectors

All the above algorithms are based on covariance matrix’s eigenvalue, and the performance is decided by the eigenvalue capability in dividing signal eigenvalue and noise eigenvalue. However, in case of limited snapshot data, the division process is quite difficult, especially at low SNR. The solution is to change eigenvalue, in which a reconstructed eigenvalue cluster at low SNR is used to extract coherent and non-coherent signal in BEM criterion [7, 8]. The performance of eigenvalue is depressed seriously by various noises, and the eigenvector is also used to replace eigenvalue to estimate source number [10].