A systolic architecture for gradient based adaptive subspace tracking algorithms

Subspace estimation plays an important role in a variety of modern signal processing applications. In an unknown and possibly changing environment, adaptive algorithms which are computationally efficient, numerically stable, and easy implementable in hardware are highly desirable. The author shows that gradient type adaptive algorithms are not only competitive in subspace tracking, but also advantageous in both the computational complexity and performance robustness. A novel systolic architecture for implementing these algorithms, with a high processor utilization, is presented