Principal components inverse: Comparisons and applications for rapidly adaptive detection

Brian Edward Freburger, University of Rhode Island

Abstract

The method of Principal Components Inverse (PCI) has been used by many researchers to address the need for rapid adaptation. More recently, variations on the idea of rank reduction have appeared in the literature. This thesis compares PCI to some of these more recent methods and applies PCI to current problems of interest in signal detection. ^ A comparison is made between applying the PCI method in element space or first transforming to beamspace and utilizing the Generalized Sidelobe Canceler (GSLC) structure for rank reduction. When the interference to be suppressed is only in the sidelobes of the mainbeam, the two methods are shown to have essentially identical performance. However, in the case of mainbeam jamming, the two methods have a departure in performance. ^ A comparison is made between PCI for the GSLC structure to two recently introduced alternatives of rank reduction for adaptive detection; the Cross Spectral Metric (CSM) and the Multistage Wiener Filter (MWF). By gaining insight into the parameters that each method utilizes and the estimation characteristics of those parameters, one can predict how each method will perform under differing scenarios. In the case of signal free training data, PCI outperforms CSM and the MWF in scenarios where the interference has a low correlation with the desired channel. In the case when there is no signal free training set, PCI outperforms CSM and the MWF due to perturbations in the correlation estimates caused by the presence of the desired signal. ^ The PCI method for interference suppression is extended to a space-time canceler structure (Vectorized Space-Time PCI) and applied to real data collected as part of the Navy's Mountaintop Radar program for the analysis of Terrain Scattered Interference. The case considered is one in which the methods are trained on the same data interval within which the associated weight vector will be applied. The VST structure is shown to provide superior cancellation ability than the currently proposed Beamspace Canceler. This increase in performance is shown to be result of including time taps in the path of the desired signal. ^

Subject Area

Engineering, Electronics and Electrical

Recommended Citation

Brian Edward Freburger, "Principal components inverse: Comparisons and applications for rapidly adaptive detection" (2000). Dissertations and Master's Theses (Campus Access). Paper AAI3025540.
http://digitalcommons.uri.edu/dissertations/AAI3025540

Share

COinS