Bayesian adaptive direction detector in sample‐starved environment

Abstract

Here, the problem of direction detection in disturbance with unknown covariance matrix is considered. The case that the number of the training data is too small to form an effective estimate for the unknown covariance matrix is focused upon. To solve the problem, the Bayesian framework is resorted to. Precisely, the unknown covariance matrix is assumed to be ruled by an inverse Wishart distribution. Then the problem is solved by the detector design criterion of generalized likelihood ratio test. Numerical examples indicate that the proposed detector can effectively detect the target, and provide higher probability of detection than the existing detectors even when sufficient training data are available.