Adaptive algorithms for array processing

This Thesis develops algorithms for the processing of data from an array of sensors. Of particular interest are techniques which may be used when the signal wavelength is much larger than the array size. The work concentrates on devising algorithms which give a minimum mean square error estimate of the signal from a specific direction in the environment. The previous algorithms under this criterion require a priori knowledge of the signal or its statistics to be available. This is rarely true in practice. The new algorithms require no a priori signal information.Path frequency domain and time domain minima mean square error processors are developed. Signal models are proposed which include all the signals reaching the array from the environment for both these cases. The frequency domain methods use complex valued data. A new algorithm which directly estimates the signal power in each direction is proposed for the frequency domain case. For the time domain case, algorithms for adapting Kalman filters and multidimensional transverse filters are derived. All of the methods proposed are designed to allow real time computations. Hence excessive numbers of multiplies,tions are avoided wherever possible. The algorithms for adapting a Adman filter require an examination of the innovations process. Similar properties of the process are used to develop autoregressive models for data from an array of sensors. These methods are related to spectral estimation techniques which are already becoming established.The relationships between several optimal processors are examined to extend the applicability of the methods developed. Expressions are obtained which are useful for realising the processors. Technique for beam forming a circular array of sensors is developed. The method is based on the Past Fourier Transform algorithm and is most suited to arrays with many sensors.The algorithms presented are supported by the results of simulation studies.