Support vector minimum variance distortionless response DOA detector

We introduce a SVM formulation of the minimum variance distortionless method (SVM-MVDM) for DOA detection. The approach can be developed in a reproducing kernel Hilbert space, thus leading to a nonlinear counterpart of the SVM-MVDM. We apply here the complex SVM formulation, and the robust cost function derivation of the SVM, and we generalize the obtained results to spectrum estimation. The classical MVDR spectrum estimator is a particular case of our SVM algorithm, though the proof of this sentence lays out of the scope of this paper.