Designing a covariance matrix tapered MVDR beamformer that is universal over notch width

Adaptive beamformers suppress interferers and reduce background noise by adjusting the complex array weights in response to the received array data. Practical adaptive beamformers like the minimum variance distortionless response (MVDR) beamformer balance these two competing requirements while maintaining unity gain for the desired look direction. The MVDR beamformer places sharp notches in the location of the interferers to minimize the interferer output power. For stationary sources, MVDR is an optimal beamformer, but the performance of MVDR degrades in the presence of moving interferers. Interferers moving at different bearing rates reside inside beamformer resolution cells for different durations, challenging MVDR’s ability to place accurate notches in the interferer direction. Consequently, the moving interferer is generally no longer within the single sharp notch location. Covariance Matrix Taper (CMT) MVDR mitigates moving interferers by creating wider notches in the beampattern. However, the CMT increases the notch width by a fixed amount, and the best notch width depends on the unknown bearing rate of each interferer which may change over time. A single fixed CMT notch width cannot suppress all moving interferers perfectly. Therefore, the need for different notch widths for different bearing rates leads us to the possibility of designing a universal algorithm for the notch width parameter. The universal CMT beamformer asymptotically achieves performance rivaling or exceeding the performance of the best fixed notch width CMT beamformer in a set by computing its array weights as a performance weighted blend of the array weights for the fixed notch width beamformer.