IMPROVED PARALLEL REALIZATION OF IIR ADAPTIVE FILTERS

A frequency-domain adaptive-filter realisation which uses a parallel arrangement of real second-order IIR sections and preprocessing based on a real orthogonal transform is examined. Experimental results and simulations show that the proposed scheme leads to increased convergence rates and lower computational complexity. The paper also deals with an efficient technique for optimising the convergence factor for the class of Gauss-Newton algorithms. Some experimental results obtained show that this technique leads to a further increase in the rate of convergence, to improved robustness of the adaptation algorithm, and to reduced mean-square error.