A frequency-domain adaptive feedback cancellation algorithm based on convex combination

In adaptive feedback cancellation systems, the partitioned block frequency-domain adaptive filter (PBFDAF) is commonly adopted to improve the computational efficiency and convergence rate while introduce a small inherent delay. However, the PBFDAF with a fixed step-size will compromise among fast convergence speed and low steady-state misalignment. In this paper, a frequency dependent convex combination of two PBFDAFs operating with different step sizes is applied, in order to achieve both fast convergence speed and low steady-state misalignment. Moreover, we design a weight transfer procedure to further improve the overall convergence performance by transferring a portion of weights of the fast filter to the slow one. Simulations results demonstrate that the proposed algorithm outperforms the affine combinaiton based method.