Robust ML estimation for unknown numbers of signals

The number of signals plays an important role in array processing. The performance of direction finding algorithms relies strongly on a correctly specified number of signals. When the number of signals is unknown, conventional approaches apply the information theoretic criterion or multiple tests to estimate the number of signals and parameters of interest simultaneously. These methods usually require a series of maximizations over parameter spaces of different dimensions and result in high computational cost. In this work, we develop a novel approach for finding ML estimates without knowing the number of signals. Given an upper bound on the number of signals, the proposed method carries out one maximization and selects relevant components from the estimated parameter vector. Simulation results show that the proposed method provides comparable estimation accuracy as the standard ML method does.