Source masking occurs when loud interferers produce high side lobes in the conventional beamformer (CBF) scanned response, obscuring the true direction-of-arrival of the desired signal. The minimum variance distortionless response (MVDR) adaptive beamformer (ABF) places deep beampattern notches near interferer directions to suppress their power in the ABF output while maintaining unity gain in the look direction. In practice, the sample matrix inversion (SMI) MVDR ABF replaces the ensemble covariance matrix with the sample covariance matrix (SCM) when computing the array weights. Estimation of the SCM in snapshot-limited and snapshot-deficient scenarios causes perturbations of the array polynomial zeros from the unit circle. The unit circle (UC) MVDR ABF projects the array polynomial zeros radially back to the unit circle, producing deeper notches and reduced sidelobe levels [Tuladhar and Buck, ICASSP 2015]. Preliminary experiments in an outdoor, free-field environment with a 21-microphone uniform linear array at a design frequency of 2.3 kHz found the UC MVDR ABF consistently outperformed both the conventional and SMI MVDR beamformers in interferer suppression when limited to 21 snapshots (one snapshot/sensor). Additionally, the UC MVDR beamformer averaged 18 dB better white noise gain than the SMI beamformer across six independent trials. [Research funded by ONR.]
Read full abstract