Alphabet CS for inverse scattering

An Alphabet Compressive Sensing (CS) approach is proposed in this paper for the retrieval of arbitrarily-shaped targets when no a-priori information on the class of scatterers at hand is available. The approach is based on the combination of (i) a fast CS retrieval methodology that is employed to invert the field data assuming several different candidate expansion bases (i.e., the alphabet), and (ii) a robust and effective algorithm for the (non-supervised) selection of the best reconstruction among those obtained with the available alphabet. Such a strategy allows to minimize the a-priori information necessary to obtain sparse representations of the unknown targets A preliminary numerical experiment is reported to validate the proposed methodology.