Multi-dimensional EEF Criterion for Source Number Estimation

When the number of snapshots is smaller than the dimension of the measured data, Exponential embedded family (EEF) rule fails to choose the correct model order. It is modified in this paper to enumerate sources in three-dimensional space unlimited by the number of snapshots. Then the modified EEF criterion is extended to R-D version for the multi-dimensional data model based on Higher-order singular value decomposition (HOSVD). The R-D EEF criterion exploits the multi-dimensional structure of measurements and the eigenvalues of different mode sample covariance matrices jointly. It improves accuracy and robustness compared with the modified EEF criterion. Simulation results demonstrate the performance of the proposed enumerators.