Parametric dictionary learning for sparsity-based TWRI in multipath environments

Abstract

Sparsity-based multipath exploitation is a promising method to eliminate ghost targets in through-the-wall radar images and utilize the additional energy in secondary reflections. The applicability of existing methods, however, is limited due to the assumption of perfectly known geometry of building interiors. We develop a parametrized multipath signal model that captures unknown or partially known wall locations. This model is used in the proposed joint image reconstruction and wall position estimation method. In order to further improve practicability in realistic scenarios, a reconstruction method based on deployment of multiple small aperture radar modules is discussed. To this end, we analyze theoretical performance bounds for colocated and distributed placements of the various modules. Supporting results based on simulated and experimental lab data are provided.