The USC Andrew and Erna Viterbi School of Engineering USC Signal and Image Processing Institute USC Ming Hsieh Department of Electrical Engineering University of Southern California

Technical Report USC-SIPI-310

“Source Localization Using Recursively Applied and Projected (RAP) MUSIC”

by John C. Mosher and Richard M. Leahy

May 1997

A new method for source localization is described that is based on a modification of the well known multiple signal classification (MUSIC) algorithm. In classical MUSIC, errors in the estimate of the signal subspace can make it difficult to accurately locate multiple sources using projections of the array vectors onto the signal subspace. Instead, recursively applied and projected (RAP) MUSIC finds multiple sources in a recursive fashion, by projecting both the signal subspace estimate and the array vectors against the orthogonal complement of the array gain matrix corresponding to the sources already found. Special assumptions about the array manifold structure, such as Vandermonde or shift invariance, are not required. We show through Monte-Carlo trials that this approach can provide improved performance in comparison to MUSIC and to the previously proposed "sequential" methods, S- and IES-MUSIC. This new method is described in the context of principal angles or principal correlations. Furthermore, through the use of these "subspace" correlations, we present a unified description that includes weighted subspace fitting methods. Finally, we describe extensions of RAP-MUSIC to cases of several sources which are diversely polarized or other vector sources which produce multidimensional array manifolds.

To download the report in PDF format click here: USC-SIPI-310.pdf (5.7Mb)