A framework for robust spectrum estimation

Beamforming and spectral estimation techniques are classified as either non-adaptive or adaptive. Non-adaptive techniques such as matched filtering provide robustness to mismatches between the assumed model and the measured data, but are susceptible to sidelobe jammers. Adaptive techniques, such as linearly constrained minimum variance beamforming, can cancel sidelobes but lack robustness to model mismatches. Further, they may perform poorly when the number of snapshots is small. To bridge the gap between non-adaptive and adaptive techniques, a new spectral estimation framework is proposed. The behavior of each estimator is controlled by two scalar-valued weighting functions. Examples of these functions yielding several popular estimation techniques are given. Methods are then developed for combining the scalar functions underlying adaptive and non-adaptive techniques to allow adaptivity to be freely traded for robustness.