Efficiency and accuracy enhanced super resolved channel estimation

This paper presents a novel approach to substantially improve the efficiency as well as the accuracy of maximum likelihood based super resolved channel estimation. Ordinary maximum likelihood methods suffer from considerable computational complexity as hundreds of coupled discrete waves have to be estimated simultaneously. Standard methods are either very slow in convergence when parameters are correlated (e.g. SAGE) or suffer from estimation errors as waves are not sufficiently decoupled. This problem is solved by modifying the likelihood function by introducing windowing of both the measurement data and the corresponding model. In this way waves are fully decoupled for fairly small separations in parameter space making it possible to perform local maximization of the corresponding modified likelihood. Moreover, as the waves are decupled a corresponding reduction of complexity is possible by clipping data which is outside the coupling distance relative to the waves subject to parameter estimation. In the case of e.g. 1000 channel frequency samples a 100-fold reduction is possible.