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-233

“Localization from Near-Source Quasi-Static Electromagnetic Fields”

by John Compton Mosher

May 1993

A wide range of research has been published on the problem of estimating the parameters of electromagnetic and acoustical sources from measurements of signals measured at an array of sensors. In the quasi-static electromagnetic cases examined here, the signal variation from a point source is relatively slow with respect to the signal propagation and the spacing of the array of sensors. As such, the location of the point sources can only be determined from the spatial diversity of the received signal across the array. The inverse source localization problem is complicated by unknown model order and strong local minima.

The nonlinear optimization problem is posed for solving for the parameters of the quasi-static source model. The transient nature of the sources can be exploited to allow subspace approaches to separate out the signal portion of the spatial correlation matrix. Decomposition techniques are examined for improved processing, and an adaptation of MUltiple SIgnal Characterization (MUSIC) is presented for solving the source localization problem. Recent results on calculating the Cramer-Rao error lower bounds are extended to the multidimensional problem here.

This thesis focuses on the problem of source localization in magnetoencephalography (MEG), with a secondary application to thunderstorm source localization. Comparisons are also made between MEG and its electrical equivalent, electro-encephalography (EEG). The error lower bounds are examined in detail for several MEG and EEG configurations, as well as localizing thunderstorm cells over Cape Canaveral and Kennedy Space Center. Time-eigenspectrum is introduced as a parsing technique for improving the performance of the optimization problem.

To download the report in PDF format click here: USC-SIPI-233.pdf (6.6Mb)