Variable step-size matrix for the improved multiband-structured subband adaptive filter

The improved multiband-structured subband adaptive filter (IMSAF) algorithm could enhance the convergence performance of multiband-structured subband adaptive filter algorithms and affine projection. However, the original IMSAF algorithm with a fixed step-size factor have to compromise between convergence rate and steady-state misalignment. A new IMSAF algorithm with variable step-size matrix (VSM) is proposed to address this problem. The power of the subband a posteriori error is set to equal the power of the subband noise to deduce the formula for VSM. Two real-time noise power estimation methods are utilized to calculate the step size. Simulation results demonstrate that the new algorithms can achieve better performance both in convergence rate and steady-state misalignment than the original IMSAF algorithm with a fixed step-size factor.