Estimation of signal directions by eigenspace based methods in nonideal environments
Analyses and remedies of the non-ideal environment effects on high-resolution eigenspace methods are presented. The eigenspace methods such as MUltiple SIgnal Classification (MUSIC) and Eigenvector Rotation (ER), are shown to give unbiased signal locations and super-resolution of signals with correct assumptions about signal, noise and array characteristics. In practice, environmental unknowns affect system inputs and/or array geometry, degrading the performance if ignored. A unified approach is adopted modeling the environments as wavefront distortions, and the distortion parameters are estimated along with the signal locations. A particular scenario is used involving signals received at a wing-mounted aircraft-array. Here, distortions due to wing flexing and fuselage scattering occur. The approach also applies to other scenarios using appropriate models. The impact of wavefront distortion models on MUSIC and ER is then examined. Results concerning bias, variance, and resolution quality of the methods are presented. Particularly, a first-order bias analysis for aforementioned distortions is given with single and two uncorrelated sources, respectively. The results show that both methods give degraded performance in distortion. ER is, however, more tolerant than MUSIC to the phase distortions. ER, unlike MUSIC, is limited to uncorrelated sources. Accordingly, the effect of signal correlation on ER is analyzed, giving biased estimates. With equi-power signals, for example, the estimates oscillate around the true arrival angles as the correlation coefficient phase varies. The oscillation magnitude depends on the correlation coefficient magnitude and the signal spacing. Finally, remedies are provided to overcome the distortions. The first scheme involves forming a pointing vector that mimics the true direction vector of the array; here use is made of the prior assumed distortion mechanisms. The scheme thus accounts for both amplitude and phase errors induced in each array element. Second, a forward-backward averaging method, which was developed for handling specular multipath, is applied to the scatter problem here encountered. This scheme decouples the multipaths from the direct signals and treats them as independent arrivals. Finally, an augmented ER based on a least-squares estimation is devised to correct the correlation effect. It provides much improved results as compared to conventional ER.
Shi, Qun, "Estimation of signal directions by eigenspace based methods in nonideal environments" (1991). Dissertations available from ProQuest. AAI9125755.