Maximum Likelihood Optimization of Adaptive Asynchronous Interference Mitigation Beamformer

In asynchronous (intermittent) interference scenarios, the content of co-channel interference sources over the data interval may be different from the interferers content over the training interval, typically with extra interference sources presented over the data interval. Under such conditions, conventional adaptive beamformer designed over the training interval may lose its efficiency when applied to the data interval. In this paper, we address the problem by 1) formulating a family of the second order statistics adaptive beamformers regularized by the covariance matrix estimated over the data interval; 2) proposing a maximum likelihood methodology for optimization of the combined (mixed) covariance matrix based on maximization of a product of a likelihood ratio that checks the accuracy of the recovered training signals and a likelihood ratio on equality of the eigenvalues in complementary to the signal subspace defined over the data interval; 3) demonstrating efficiency and robustness of the proposed solution as a linear adaptive beamformer and as an initialization for iterative beamformer with projections to the finite alphabet in different asynchronous interference scenarios comparing with the basic training and data based interference rejection combining receivers.