Subspace fitting concepts in sensor array processing

The area of sensor array processing has recently received considerable attention in the litterature. Much interest stems from the potential of applying advanced concepts from signal processing to a variety of real-world problems. In these applications, spatially distributed sensors are used to detect and locate multible emitters and/or to seperate the individual encountered in communication and radar applications, when the signal transmission is subject to undesired interference. Acoustic devices (hydrophones) are used in, e.g., underwater and seismic applications.A vast number of methods have been proposed for estimating unknown parameters of the received wavefronts. In Part II of this thesis, an attempt is made to examine the relations among many of these techniques, as well as the accuracy of the resulting estimates. Several estimation methods are posed as solutions to different versions of a basic subspace fitting problem. The asymptotic (for large data records) performance of the multidimensional subspace fitting methods is investigated in Parts II, IV, and V. A new estimator, termed the Weighted Subspace Fitting (WSF) method is introduced and analysed. The WSF method is shown to be a unification of the subspace fitting techniques and the stochastic Maximum Likelihood method, valid for large amounts of data. The analysis is carried out for the case of Gaussian noise and Gaussian/nonGaussian signal waveforms.In Part III, a numerical procedure for calculating the WSF estimates is proposed. A detection scheme based on the WSF method is also suggested, and shown to yield a consistent estimate of the number of coherent/non-coherent wavefronts.Part VI deals with the problem of separating a desired signal from unwanted disturbances. The desired signal is assumed to be absent in certain time (or frequency) intervals, and the proposed method requires no calibration of the array and involves no numerical search.