On frequency-domain least squares adaptive algorithms

The simple least mean-square (LMS) algorithm shows slow convergence rate, while the computationally-intensive least squares (LS) algorithm has very fast convergence speed. The LMS algorithm realized in the frequency domain is known to have significantly improved convergence rate with moderate complexity. In this work we study new adaptive algorithms that take advantage of both the LS and frequency-domain techniques. We develop several frequency-domain LS algorithms and investigate their performance characteristics by computer simulation.