The application of linear constraints to an adaptive beamformer

This paper describes a novel technique for implementing the constrained adaptive array solution. It involves the use of a simple full-rank transformation network to pre-process the received data prior to using a canonical "end element clamped" adaptive combiner. It will be shown that the form of the transformation is nonunique and how a number of particularly efficient realisations can be devised. The resulting preprocessors only require O(N) arithmetic operations per time sample for an N channel array and can be implemented using a simple linear array of processing elements. It will also be shown how the pre-processor technique can be generalised to encompass multiple linear constraints and that these can also be implemented using efficient pre-processor networks of low computational complexity.