Efficient widely linear beamforming using oblique projection

초록

When noncircular signals are incident into a sensor array, the performance of an adaptive beamformer can be improved by introducing a widely linear (WL) scheme with an extended virtual array that is formed by concatenating the actual array with its complex conjugated version, which leads to the conjugate symmetry of the extended array. Fast convergence and robustness to steering vector errors can be attained using a reconstructed interference-plus-noise covariance matrix (INCM), rather than a sample matrix. This paper presents two WL beamforming methods through the estimation of the noncircularity of the desired signal and the INCM. Exploiting oblique projection, the proposed methods estimate the INCM without seeking noncircularities and powers of interferences. They are different in obtaining the oblique projection. One attains it without the estimation of interference directions while the other estimates them. The oblique projections are found through the eigen-decomposition of the extended sample matrix, which is efficiently made using the conjugate symmetry. Simulation results demonstrate the effectiveness of the proposed methods, which, showing robustness against steering vector errors, outperform existing WL beamformers. Moreover, they can achieve the optimum signal-to- interference-plus-noise ratio (SINR) at the steady state where the number of samples is infinite.

키워드