Emitter number detection based on clustering and the application of information theoretic criteria

An approach for determining the number of emitters for array processing is proposed based on the asymptotic distribution of eigenvalues of a sample covariance. Instead of using the set of the smallest eigenvalues, different eigenvalue groupings are used as the competing models. An information theoretic criterion is used for determining the optimal grouping. The approach is able to solve the eigenvalue correspondence problem associated with most emitter number detection algorithms and demonstrates an improved detection performance under certain circumstances. The approach is computationally more efficient than order statistics based approaches. Computer simulations are used to demonstrate the performance of the proposed approach.