Pruning sparse signal models using interference

Previous work on sparse approximations has shown that in the pursuit of a signal model using greedy iterative algorithms, the efficiency of the representation can be increased by considering the interference between selected atoms. However, in such interference-adaptive algorithms, atoms are still often selected that necessitate correction by subsequently chosen atoms. It is thus logical to remove these atoms from the representation so that they do not diminish the efficiency of the pursued signal model. In this paper, we propose to prune atoms from the model based on the degree and type of interference, and test its effectiveness in an interference-adaptive orthogonal matching pursuit algorithm.