Bias Corrected PSD Estimation for an Adaptive Array With Moving Interference

Abstract

We address the issue of computing power spectral density (PSD) estimates at the output of a beamforming sensor array in the presence of strong moving interference. It is shown that the time-varying spatial response of an adaptive beamformer (ldquopattern rumblerdquo) causes estimation bias in the PSD of both the signal of interest (SOI) and noise. In applications such as radio astronomy with stringent sensitivity requirements, even small pattern variations can be problematic because the resulting higher variance noise spectrum estimates make it impossible to detect signals of interest which are many decibels below the noise floor. Distortion in beam mainlobe shape also introduces errors in SOI direction estimates. To overcome this problem, an extension of the method described in Leshem , 2000, is developed which eliminates pattern-distortion-induced PSD bias and spatial response errors over the long-term PSD averaging window. Both simulated and real data experiments demonstrate algorithm effectiveness in realizing an undistorted effective (average) beam spatial response while maintaining a low noise floor level. This algorithm will enable PSD estimation using multi-antenna sensors and adaptive interference cancellation for radio astronomy, remote sensing, and other sensitive radiometry applications where cancellation has not been feasible.