Adaptive Detection Based on Bayesian Approach in Heterogeneous Environments
非均匀杂波环境下基于贝叶斯方法的自适应检测

The performance of adaptive detection of an interested signal degrades when the environment is heterogeneous, i.e., the training data samples used for adaption do not share the same covariance matrix as the cell under test (CUT). To circumvent the problem, a Bayesian generalized likelihood ratio test (B-GLRT) detector is derived. On the one hand, the heterogeneity is considered at the design stage of B-GLRT by means of the statistical modeling of the covariance matrixes of CUT and training data in heterogeneous environment. Meanwhile, the degree of heterogeneity can be tuned through scalar. On the other hand, a prior distribution is assigned to covariance matrix to exploit some prior knowledge for performance improvement. Numerical simulations show that B-GLRT outperforms the conversional non-Bayesian detectors. Meanwhile, the influence of heterogeneity and prior knowledge on detection performance is illustrated.