Nominal direction estimation for slightly distributed scatterers using the SAGE algorithm

Two SAGE (subspace-based alternating generalized expectation-maximization) algorithms are derived using a deterministic (D-) version and stochastic (S-) version of the generalized array manifold (GAM) model proposed by D. Asztely et al. (see Proc. ICASSP´97, 1997) for estimation of nominal azimuths of arrival (NAoAs) of multiple slightly distributed scatterers (SDSs). Monte-Carlo simulation studies in a single-SDS scenario show that the S-GAM SAGE algorithm returns lower root mean squared estimation error than both the D-GAM SAGE algorithm and the spread-ESPRIT technique based on a two-ray model (Bengtsson, M. and Ottersten, B., 2000). Furthermore, in a two-SDS scenario with strong power unbalance between the SDSs, the S-GAM SAGE algorithm demonstrates best performance in estimating the NAoA of the SDS with weakest power compared to the other estimators. The simulation results further demonstrate that when multiple slightly distributed scattering occurs, the investigated estimators can estimate the NAoA with an effective accuracy, provided the SDSs are separated by more than twice the intrinsic azimuth resolution of the array. The proposed algorithms can be easily extended to include the nominal direction, i.e. azimuth and elevation, of arrival and direction of departure of the SDS when multiple-input and multiple-output (MIMO) applications are considered.