Low-Complexity Bayesian-GLRT-Based Target Detection for MIMO Radar in Gaussian Clutter

This paper proposes an unstructured adaptive moving target detection approach for a multiple-input multiple-output (MIMO) radar to obtain reduced computational complexity. Moreover, the MIMO radar operates in a Gaussian clutter with an unknown but stochastic covariance matrix. At the design stage, the Bayesian unstructured generalized likelihood ratio test (UGLRT), namely the BUGLRT detector, which inherits from the UGLRT approach, has been derived. The BUGLRT approach has transformed 2D searching of structured generalized likelihood ratio test (SGLRT) based on the Bayesian framework, namely BSGLRT, detector into a 1D searching procedure that drastically reduces the computational complexity. The asymptotic probability of false alarms and computation complexity have also been analyzed. Both theoretical analysis and numerical simulation results show that BSGLRT and BUGLRT detectors outperform their non-Bayesian counterparts, especially for a small number of snapshots. Moreover, the detection performance of BUGLRT is better than that of BSGLRT, but the performance gap decreases as the number of snapshots increases. In addition, the detectors based on the Bayesian framework are sensitive to the existing mismatches of parameters for inverse complex Wishart distribution. Compared with BSGLRT, the BUGLRT can obtain a noticeable complexity reduction along with a slightly improved detection performance.