Performance of capacity inference methods under colored interference
Résumé
In this paper, we address the problem of fast point-to-point channel capacity estimation in the situation where the receiver undergoes unknown colored interference from multiple sources, whereas the channel with the transmitter is perfectly known. Using estimation techniques associated to large random matrix theory, we derive an estimator referred to as the G-estimator and compare its performance against the conventional estimator. In particular, we prove that, unlike the conventional estimator, the G-estimator is consistent in the large dimensional setting. We finally complete the analysis by describing its fluctuations: When properly centered and rescaled, the G-estimator satisfies a central limit theorem, hence has Gaussian fluctuations. Simulations are provided which clearly show that the G-estimator outperforms the conventional one; simulations also strongly support the theoretical results even for small system dimensions.
Origine : Fichiers produits par l'(les) auteur(s)