Frequency Domain Adaptive Filters推荐.pdf

1 • Block Adaptive Filters Iterating LMS under the assumption of small variations in w (n) Approximating the gradient by time averages The structure of the Block adaptive filter Convergence properties • Frequency Domain Adaptive Filters Frequency domain computation of linear convolution Frequency domain computation of linear correlation Fast LMS algorithm Improvement of convergence rate Unconstrained frequency domain adaptive filtering Self-orthogonalizing adaptive filters Reference: Chapter 7 from Haykin’s book 2002 2       • the (correlated) input signal samples {u(1), u(2), u(3), . . .},     generated randomely;     • the desired signal samples {d(1), d(2), d(3), . . .} correlated     with {u(1), u(2), u(3), . . .}    with an arbitrary parameter vector w (0), for example w (0) = 0. n = 0, 1, 2, 3, . . . , nmax Read /generate a new data pair, (u(n), d(n)) (Filter output) y (n) = w (n)T u(n) = ∑M w (n)u(n − i)