Near-field sources localization: a model-fitting approach

Beamforming algorithms handle sources that are located in the near field by simply adapting the steering vector and having it depend not only on bearing but also on range. The corresponding algorithm, known as the focused beamformer, has a poor performance even though the computational burden is quite important since it involves a 2-dimensional beam evaluation in the bearing-range plane. In order to localize such sources using a linear equispaced array in a narrowband context, we propose a deconvolution approach to the output of a beamformer performing a one-dimensional evaluation. For reasonably difficult scenarios, the performance we obtain is satisfactory (close to the Cramer-Rao bound) for a quite acceptable computational burden