Estimating the number of signals using reference noise samples

Criteria for estimating the number of signals in the case of small-sample populations are considered. The criteria introduced by M. Wax and T. Kailath (1985) do not work well in this case. Two new criteria based on information theoretic and eigenmode composition-based criteria are proposed, assuming a known noise covariance. Simulation results show a significant improvement in the performance of estimating the number of signals in the small-sample case, by the use of the new criteria combined with an estimate of the noise covariance obtained from reference noise samples.<>