Rapidly adaptive signal detection using the principal component inverse (PCI) method

This paper presents a review of the principal component inverse method of rapidly adaptive signal detection and contrasts this method with the cross spectral metric method of rank reduction. Rank reduction methods are most effective if (a) the vector representing the strong portion of the noise lies in a low-dimensional subspace, (b) the subspace remains nearly constant for a local set of interference vectors which is large enough for good estimation of both the dimension of the subspace and a set of basis vectors for the subspace, and (c) signal suppression is unlikely.