Two subspace methods for frequency sparse graph signals (Q6652571)

The authors investigate signals which are sparse in the graph's spectral domain. They develop explicit algorithms to not only reconstruct the support set but also partial components from samples arising from a few graph vertices. The number of required samples is independent of the total size of the graph and takes only local properties of the graph into account. The derived results are based on an operator-theoretic framework for subspace methods and are shown to become effective if the spectral eigenfunctions are zero-free or linearly independent on a small set of graph vertices.