Adaptive Radar Signal Detection with Integrated Learning and Knowledge Exploitation

We consider the problem of weak signal detection in strong disturbance with a subspace structure. Unlike conventional subspace detection techniques relying on the availability of a large amount of training data, we consider a knowledge-aided (KA) subspace detection approach for training limited scenarios by incorporating partial prior knowledge of the subspace. A unique feature of the proposed approach is that it can identify the missing subspace bases and recover the full subspace structure by using only the test signal, thus bypassing the need for training data. The proposed approach utilizes a Bayesian hierarchical model for knowledge representation. The model is integrated within a sparse Bayesian framework, which promotes parsimonious subspace representation of the observed data. A variational Bayesian inference algorithm is developed based on the proposed model to recover parameters and subspace structures associated with the disturbance, which are then brought into a generalized likelihood ratio test (GLRT) to perform signal detection. Numerical results are presented to illustrate the performance of the proposed subspace detector in comparison with several notable existing methods.