Radar sparse signal processing by non-negative least-squares estimation

Increasing the granularity of the solution space for many inverse problems results in ill-posedness. Detection of objects that are close in the solution space e.g. in range or Doppler is significantly improved if a-priori information, such as sparsity, can be used. Alternatively to traditional sparse signal processing algorithms priors, in this paper, we exploit the non-negativity of objects power estimates and propose the use of a non-negative least-squares algorithm to perform single-pulse target estimation. We discuss the corresponding gridding and objects relative-phase effects on estimation as well the relation with other sparse signal processing methods. Finally, we present numerical results supporting that the proposed method, similarly to typical sparsity based algorithms, has the potential to increase detection probability of closely spaced objects.