This dissertation deals with spatial signal processing of distributed source estimation using an antenna array where a typical example of the distributed source model results from a wireless channel with multipath local scattering. The work is primarily motivated by the promising application of antenna arrays at the base stations of a cellular system to improve the system capacity. Specifically, the problems covered in this dissertation are mainly related to the description of spatially distributed sources and direction of arrival (DOA) estimation of the distributed sources. Two main topics studied in depth are DOA estimation of coherently distributed sources and that of incoherently distributed sources.
For a coherently distributed source model, we consider the estimation of two-dimensional (azimuth and elevation) DOA using a pair of uniform circular arrays. We propose a low-complexity estimation algorithm, called the sequential one-dimensional (SOS) searching algorithm by concentrating only on the estimation of DOA`s. The SOS algorithm has a basis on the eigenstructure between the steering matrix and signal subspace, and utilizes a preliminary estimate. The proposed algorithm exhibits as good an estimation performance as the maximum likelihood method for coherently distributed sources.
For an incoherently distributed source model, we consider the problem of estimating the DOA`s. The performance of most subspace-based high-resolution DOA estimation algorithms degrades and ML-based DOA estimation techniques require a high-dimensional nonlinear optimization problem, when the spatial covariance matrix has full-rank as in the case of an incoherently distributed source. We propose a novel estimation method based on the conventional beamforming approach, which estimates the DOA from a spatial maximum peak of the output power: the proposed method is on the average computationally more attractive than the conventional ML-based methods and provides better performance...