Nominal Direction and Direction Spread Estimation for Slightly Distributed Scatterers using the SAGE Algorithm

Xuefeng Yin, B.H. Fleury
2005 IEEE 61st Vehicular Technology Conference  
In this paper, the SAGE (Subspace-Alternating Generalized Expectation-maximization) algorithm [1] [2] is derived using the generalized array manifold (GAM) model proposed in [3] (GAM-SAGE) to estimate the nominal directions, i.e. azimuths and elevations of slightly distributed scatterers (SDSs). As byproducts estimates of the azimuth spreads (ASs), elevation spreads (ESs), and the azimuth-elevation correlation coefficients (AECCs) of the SDSs can be computed from the estimates of the GAM
more » ... s of the GAM parameters. These parameters determine with close accuracy the direction spreads [4] of SDSs. Simulation studies show that in a single-SDS scenario, the GAM-SAGE algorithm outperforms the Spread-ESPRIT technique, and both of them outperform the SAGE algorithm derived with the conventional specularscatterer (SS) model (SS-SAGE) when the output signalto-noise ratio (SNR) is beyond a certain threshold which depends on the AS and ES of the SDS. In a two-SDS scenario with strong power unbalance between the SDSs, provided the direction spacing between the SDSs equals twice the intrinsic azimuth or elevation resolution of the array, the GAM-SAGE algorithm can estimate the nominal direction of the SDS with weakest power with tolerably small errors. The SS-SAGE algorithm returns high root mean squared estimation error (RMSEE) regardless of the direction separation. We also found that the AECC estimator needs to operate in high SNR in order for its bias and RMSEE to be tolerably small. The performance of the AECC estimator, as well as the AS and ES estimators can be improved by applying an array-size selection technique proposed in [5].
doi:10.1109/vetecs.2005.1543242 fatcat:zgldkdbrhzbgzg6lljuhjvdfgi