Algorithms and architectures for adaptive array signal processing

Antenna arrays sample propagating waves at multiple locations. They are employed e.g. in radar, sonar and wireless communication systems because of their capacity of spatial selectivity and localization of radiating sources. Current model-based algorithms make use of computationally demanding orthogonal matrix decompositions such as the singular value decomposition (SVD). On the other hand the data rates are often extremely high. Therefore, real-time execution of complex algorithms often requires parallel computing. We study the simultaneous design of new algorithms and parallel architectures for subspace tracking, for robust adaptive beamforming and for direction finding of narrow-band and wide-band sources. By structuring all recursive algorithms in a similar way, they can be mapped efficiently onto the Jacobi architecture, which was originally developed for SVD updating. The numerical and architectural aspects of this algorithm are improved by the use of a minimal parameterziation of the orthogonal matrix of short singular vectors. Finally, a new Fourier-based linear model for direction finding in colored ambient noise fields is proposed.