On detection of the number of signals when the noise covariance matrix is arbitrary (Q1822171)

The authors propose some model selection methods for determination of the number of signals in presence of noise with arbitrary covariance matrix. This problem is related to finding the multiplicity of the smallest eigenvalue of \(\Sigma_ 2\Sigma_ 1^{-1}\), where \(\Sigma_ 2=\Gamma +\lambda \Sigma_ 1\), \(\Sigma_ 1\) and \(\Sigma_ 2\) are covariance matrices, \(\lambda\) is a scalar, and \(\Gamma\) is an non-negative definite matrix and is not of full rank. By means of some information theoretic criteria some methods for determination of the multiplicities of various eigenvalues of \(\Sigma_ 2\Sigma_ 1^{-1}\) are proposed. The likelihood ratio test and other test procedures for testing the hypothesis that the last few eigenvalues of \(\Sigma_ 2\Sigma_ 1^{-1}\) are equal to \(\lambda\) (\(\lambda\) is known or unknown) are discussed. The strong consistency of these criteria is established.