High resolution algorithms for array signal processing

This thesis considers the performance of the existing and the development of the new high resolution algorithms, which are applied to array signal processing problems. The performance prediction of the MUSIC and Whitened MUSIC algorithms are investigated, where a second-order and a first-order perturbation analysis are performed for the white noise and coloured background noise cases. The predictions spectra have given good agreement with the simulated results. We introduce a family of different beamformers from a general purpose matrix, such that MUSIC and Capon's MLM are from the same family of the methods. We have obtained a single root-location algorithm which has resulted on a common variance expression for the root-locations and the root-MUSIC algorithm achieves the minimum variance in this family. Maximum likelihood methods (MLM) are known to perform well, however their high computational cost is reduced using a new suboptimal MLM algorithm for uncorrelated sources. This achieves good performance at low SNR, in comparison with MUSIC. It is then extended to the highly or fully correlated sources problem. The new MUSIC-Type algorithms are investigated for uncorrelated signals and a new iterative general purpose algorithm based on this is derived. The algorithms are shown to perform well at a wide variation of SNR's. The new MUSIC methods can be applied with other established techniques such as spatial smoothing or CSSM to yield good performance in the fully correlated case.