Information-Theoretic Approaches to Radar Target Enumeration

When the number of antennas is comparable to the number of snapshots, the classical source enumerators cannot correctly detect the number of signals impinging upon an array. To tackle this issue, we develop a linear shrinkage-based minimum description length (LS-MDL) criterion and two shrinkage coefficient-based detectors (SCDs) for source enumeration. In the LS-MDL approach, the sample covariance matrix of the noise subspace components can be accurately calculated, of which eigenvalues are used to determine LS-MDL that is proved to be of strong consistency for m, n → ∞ and m/n → c ε (0, ∞) where m and n are the numbers of antennas and snapshots, respectively. Moreover, we prove that the noise shrinkage coefficients are asymptotically Gaussian distributed as m, n → ∞ and m/n → c. Utilizing the properties, the threshold-like and heuristic SCDs are devised. Simulation results are included for illustrating the effectiveness of the proposed methodologies. © 2025 by The Institute of Electrical and Electronics Engineers, Inc.