Spatial-spectrum estimation using vectors closest to the array manifold

Consideration is given to multiple narrowband source localization using arbitrarily configured arrays and spatial-spectrum estimation. A new eigenspace based approach is described which employs projections onto a particular vector in the estimated noise-only subspace. The vector is one which is, in some sense, closest to the section of the array manifold corresponding to a source location sector of interest. This approach incorporates a priori knowledge of the array manifold over a location sector of interest to provide signal-to-noise ratio (SNR) spectral-resolution thresholds and location estimation variances which are lower than those of the MUSIC and MIN-NORM methods. For some arrays, these can be substantially lower. Several CLOSEST vector estimators are developed by employing different measures of closeness. Simulations comparing CLOSEST with MUSIC and MIN-NORM methods are presented. The results indicate that CLOSEST vectors can, for some arrays, perform significantly better than MIN-NORM vectors