Multi-dimensional DFT beamspace SAGE super-resolution algorithm

A novel multi-dimensional frequency domain SAGE (space-alternating generalised expectation-maximisation) algorithm in DFT beamspace for estimating channel parameters is presented. In contrast to the classical implementation of the SAGE algorithm in element space, the proposed algorithm allows the estimates of each parameter to be optimised within a smaller domain. Further reduction in the computational complexity within each SAGE iteration is achieved since the alternating parameters in the correlation process are in the real-valued domain. A major advantage of the proposed algorithm is its ability for mapping onto parallel processors. The general implementation of the proposed algorithm with efficient matrix computation is presented, with conditions necessary for achieving the best performance stated. Finally, its performance is evaluated by using real channel measurement data.